Geographic Access Barrier as a Critical Mediator in Obstetric Fistula Treatment Cascade in Nigeria: Evidence from a Causal Mediation Analysis of 5,496 Cases

Abstract

Background

Approximately two million women globally currently live with untreated obstetric fistula, predominantly in sub-Saharan Africa and South Asia, with Nigeria accounting for 40% of cases. Despite the recognition of multiple healthcare barriers, the causal mechanisms through which these barriers affect treatment-seeking behavior remain poorly understood. We conducted a causal mediation analysis to examine potential pathways in the fistula treatment cascade.

Methods

We analyzed data from 5,496 women who reported fistula symptoms in the 2024 Nigeria Demographic and Health Survey. Using the counterfactual framework for causal mediation, we decomposed the relationship between fistula knowledge (exposure) and treatment-seeking (outcome) into direct and indirect effects operating through three mediator domains: financial, geographic access, and autonomy barriers. We estimated natural direct and indirect effects using Baron and Kenny regression with product-of-coefficients method, inverse probability weighting, and doubly robust estimation. Bootstrap confidence intervals (1,000 replications) and E-value sensitivity analyses were used to assess the robustness of sampling variability and unmeasured confounding.

Results

Among women with fistula, 65.0% sought treatment, 3.3% received surgical repair, and 6.9% achieved resolution leaving 93.1% with unresolved fistula. The largest cascade gap (61.7%) occurred between treatment-seeking and surgery receipt. Geographic access barriers significantly reduced treatment-seeking (OR=0.73; 95% CI: 0.59–0.88; p=0.002), emerged as the only significant mediating pathway (indirect effect β=−0.19; 95% CI: −0.33 to −0.07). Financial barriers (OR=1.11; 95% CI: 0.95–1.31) and autonomy barriers (OR=0.99; 95% CI: 0.84–1.18) were not independently associated with treatment-seeking in adjusted models. The E-value for geographic access barriers was 2.08, indicating moderate robustness to unmeasured confounding.

Conclusion

Geographic access barriers emerged as the dominant pathway associated with reduced treatment engagement, findings consistent with these barriers suggesting a critical pathway under stated causal assumptions. Although knowledge was not directly associated with treatment-seeking, mediation models suggested that geographic access barriers were the only pathway consistently linked to care engagement. Financial and autonomy-related barriers were not statistically significant mediators in this analysis, though this may reflect measurement limitations or operation at different cascade stages rather than true irrelevance. The primary cascade bottleneck is surgical access rather than treatment-seeking, suggesting health system capacity constraints. These findings align with a causal interpretation under the stated identification assumptions of the counterfactual framework, though temporal ordering cannot be definitively established from cross-sectional data. Interventions should prioritize mobile surgical outreach, transportation vouchers, and decentralizing fistula repair services.

Plain Language Summary

Obstetric fistula is a childbirth injury that causes women to uncontrollably leak urine or feces, often leading to social isolation, divorce, and severe psychological distress. Although surgery can repair this condition, most affected women in Nigeria and other low-resource countries never seek treatment, even when they know that treatment exists. Our study sought to understand why this gap between knowledge and action persists. We analyzed survey data from nearly 5,500 Nigerian women to examine whether the reason women do not seek treatment is that knowing about fistula also makes them more aware of the barriers standing in their way, such as the cost of care, difficulty reaching a hospital, or lack of permission from family members. We found that women who knew more about fistulas were more likely to perceive all three types of barriers; however, only access barriers, such as distance to health facilities and lack of transportation, actually prevented them from seeking care. Financial concerns and the need for family permission did not significantly reduce the treatment-seeking behavior. This indicates that simply educating women about fistulas is insufficient; health programs must also address the practical challenges women face in physically reaching treatment centers. Bringing fistula repair services closer to affected communities through outreach camps, mobile surgical units, or decentralized care models may be more effective than awareness campaigns alone in helping women access the life-changing surgery they need.

Introduction

Obstetric fistula, an abnormal opening between the vagina and bladder or rectum resulting from prolonged obstructed labor, represents one of the most severe and neglected maternal morbidities globally.^1,2 The condition affects approximately two million women worldwide, with an estimated 50,000–100,000 new cases occurring annually, predominantly in sub-Saharan Africa and South Asia.^2,3 Nigeria bears a disproportionate burden, accounting for approximately 40% of global fistula cases, with over 200,000 women currently living with the condition and an estimated 12,000 new cases occurring each year.⁴ The consequences of obstetric fistula extend far beyond the physical symptoms of urinary or fecal incontinence. Affected women frequently experience social ostracism, marital dissolution, economic marginalization, and profound psychological distress, including depression and suicidal ideation.^5,6 Despite being surgically correctable in most cases, with success rates exceeding 90% at specialized centers, the vast majority of women with fistula never receive definitive treatment.⁷ This treatment gap has been characterized as a failure of both the healthcare system and broader social structures, reflecting deep-seated inequities in access to maternal health services.⁸

Previous research has identified multiple barriers to fistula treatment, including financial constraints, geographic inaccessibility, lack of awareness about treatment availability, stigma and shame, limited decision-making autonomy, and insufficient surgical capacity.^9,10 Nigeria’s health system faces substantial constraints in fistula surgical capacity. Specialized fistula repair requires skilled surgeons, often with additional training in obstetric and gynecologic reconstruction, typically requiring 6–8 weeks of competency-based training in accredited facilities.^11–13 Such expertise is concentrated in a small number of tertiary referral hospitals and specialized fistula centers, primarily located in urban areas.¹⁴ Estimates suggest Nigeria has approximately 30–50 surgeons with dedicated fistula training serving a population exceeding 200 million, creating vast geographic inequities in access.^15,16 Referral systems are frequently inefficient, with breakdowns in communication, coordination, and follow-up between primary health centers and surgical facilities.^17,18 These system-level constraints mean that even women who successfully navigate individual-level barriers to seek treatment may encounter insurmountable obstacles in accessing definitive surgical repair.¹¹ Systematic reviews have documented that these barriers operate at individual, community, and health system levels, often interacting in complex ways to prevent women from accessing care.¹⁰ However, most existing studies have employed descriptive or cross-sectional associational methods, limiting our understanding of the causal mechanisms through which specific barriers affect treatment-seeking behaviors.

Fistula knowledge was selected as the primary exposure because awareness represents the most common target of public health interventions. Governments and non-governmental organizations have invested substantially in awareness campaigns, radio announcements, community health education, and stigma reduction programs predicated on the assumption that knowledge facilitates care-seeking.^11,18,19 However, these programs have not yielded proportional increases in treatment uptake,^4,11,20 suggesting that the knowledge-to-treatment pathway may be more complex than direct facilitation. Understanding whether and how knowledge influences treatment-seeking, and through which specific barrier mechanisms, is essential for optimizing intervention design and resource allocation. An important conceptual consideration is that awareness may operate through complex mechanisms. The Health Belief Model posits that knowledge represents a necessary but insufficient precursor to health-seeking behavior, with perceived benefits and perceived barriers mediating the relationship between knowledge and action.²¹ While health education is presumed to facilitate care-seeking by reducing information barriers, knowledge may simultaneously heighten recognition of structural constraints^22–24 - that is, informed women may be more realistic about the obstacles they face. This bidirectional relationship between knowledge and perceived barriers requires explicit consideration in mediation models.

Causal mediation analysis provides a rigorous framework for decomposing total effects into direct and indirect pathways, enabling the identification of the mechanisms through which exposures influence outcomes.²⁵ The counterfactual approach to mediation, developed over the past two decades, offers several advantages over traditional regression-based methods, including explicit articulation of causal assumptions, accommodation of exposure-mediator interactions, and well-defined effect decomposition even with binary outcomes.²⁶ The application of these methods to obstetric fistula has been notably absent from the literature, representing a critical gap in evidence needed to inform intervention design.

Understanding the treatment cascade for obstetric fistula from symptom onset through treatment-seeking, surgical repair, and resolution is essential for identifying intervention targets. The cascade framework, widely applied in HIV care to track patients from diagnosis through viral suppression, enables systematic identification of the stages in the care continuum where the greatest attrition occurs.²⁷ For obstetric fistula, the cascade differs in important ways: unlike HIV, where the endpoint is ongoing viral suppression, the fistula cascade has a definitive endpoint (surgical cure and symptom resolution). Additionally, while HIV diagnosis typically occurs at a single moment, fistula symptoms may be persistent, intermittent, or may vary in severity, complicating recognition and treatment-seeking timing. Nevertheless, the cascade model provides a structured approach to quantifying gaps in care and prioritizing interventions at the stages with highest attrition.²⁸

A methodological consideration for obstetric fistula epidemiology is the reliance on self-reported symptoms in large-scale surveys. Clinical examination would provide gold-standard confirmation, but is infeasible in population-based surveys covering tens of thousands of women across geographically dispersed areas. Validation studies have demonstrated that DHS fistula symptom questions have acceptable sensitivity (72–89%) and specificity (89–95%) when compared against clinical diagnosis by gynecologists, though false positives can occur with other causes of incontinence. This measurement limitation is important to acknowledge, though it affects prevalence estimation more substantially than the examination of associations between barriers and treatment-seeking behaviors.

In this study, we applied causal mediation analysis using the counterfactual framework to data from the 2024 Nigeria Demographic and Health Survey (NDHS), the largest nationally representative survey that includes comprehensive fistula-related questions. Our objectives were threefold: (1) to characterize the treatment cascade for obstetric fistula at a national scale; (2) to identify which healthcare barrier domains (financial, geographic access, or autonomy) mediate the relationship between fistula knowledge and treatment-seeking; and (3) to quantify the direct and indirect effects of knowledge on treatment-seeking through each barrier pathway. We hypothesized that geographic access barriers would emerge as the primary modifiable mediators, given Nigeria’s vast geography, limited transportation infrastructure, and concentration of fistula surgical expertise at tertiary centers.

Methods

Study Design and Data Source

We conducted a cross-sectional analysis of the 2024 Nigeria Demographic and Health Survey (NDHS), a nationally representative household survey implemented by the National Population Commission (NPC) with technical assistance from ICF International through The DHS Program.²⁹ Data collection occurred from December 2023 to May 2024, employing a stratified two-stage cluster sampling design. The sampling frame comprised 42,000 households across 1,400 enumeration areas (primary sampling units) selected from all 36 states and the Federal Capital Territory, stratified by urban and rural residence within each state. The survey achieved a 99% household response rate and 97% individual response rate among eligible women aged 15–49 years. All women aged 15–49 years who were either permanent residents or visitors present in selected households on the night preceding the survey were eligible for individual interviews. The women’s questionnaire included a fistula module that assessed symptom experience, treatment-seeking behavior, surgical history, and current fistula status. Healthcare barrier questions assessed difficulties in accessing care across multiple domains.

Study Population

Of the 39,050 women who completed the individual interview, 23,474 (60.1%) responded to the fistula symptom questions and constituted our analytical sample for descriptive comparisons. The fistula module was administered only to women who had ever been pregnant, as obstetric fistula results from childbirth complications. Among all surveyed women, 15,576 (39.9%) had never been pregnant and were therefore not asked fistula-related questions. This sample restriction was by design and does not represent non-response or missing data among eligible participants. Among respondents, 5,496 women (23.4%) reported experiencing symptoms consistent with obstetric fistula (involuntary leakage of urine or feces from the vagina following childbirth). These women comprised the primary analytical sample for mediation analyses. A complete case analysis was performed on 4,236 women with non-missing data on all variables (77.1% of fistula cases).

Measures

Exposure

Fistula knowledge was assessed through a binary indicator of whether the respondent had ever heard of fistula as a health condition (yes/no). This variable captured awareness of fistula as a treatable condition, which was hypothesized to influence treatment-seeking through both direct and indirect pathways operating through healthcare barrier perceptions.

Outcome

The primary outcome was treatment-seeking for fistula, defined as a binary indicator of whether the respondent had ever sought medical care for fistula symptoms (yes/no). The secondary outcomes included receipt of surgical repair (yes/no) and fistula resolution (yes/no).

Mediators

Healthcare barriers were assessed through eight items asking whether each factor represented a “big problem” in accessing healthcare: (1) getting money for treatment, (2) distance to health facility, (3) having to take transportation, (4) getting permission to go, (5) not wanting to go alone, (6) concern about female provider availability, (7) concern about drug availability, and (8) concern about provider availability. We constructed three composite mediator variables: financial barriers (money as a big problem), geographic access barriers (distance or transportation as a big problem), and autonomy barriers (permission or not wanting to go alone as a big problem). Each composite was coded as binary (any barrier present vs. no barrier present). We constructed three composite mediator variables using binary indicators (any barrier present vs. none) to facilitate interpretability in logistic mediation models and to avoid assuming linear dose-response relationships. While this approach may attenuate effect estimates by collapsing gradients of barrier severity, it provides conservative estimates and aligns with policy-relevant binary classifications of barrier presence.

Statistical Analysis

We calculated frequencies and proportions for categorical variables and means with standard deviations for continuous variables, stratified by fistula status. Chi-square tests and independent-sample t-tests were used to assess the differences between the groups. The treatment cascade was characterized by calculating the proportion of fistula cases at each stage: (1) total with symptoms, (2) sought treatment, (3) received surgical repair, and (4) achieved resolution. Attrition at each stage was quantified as the absolute and proportional gaps from the preceding stage.

Causal Mediation Framework

We employed the counterfactual framework for causal mediation analysis, as described by VanderWeele.^30,31 All mediation models were adjusted for potential confounders which includes maternal age, education level, household wealth quintile, urban/rural residence, geopolitical region, religious affiliation, early marriage (marriage before age 18), and parity. Let A denote exposure (fistula knowledge), M the mediator (healthcare barrier), Y the outcome (treatment-seeking), and C a vector of baseline covariates. All mediation models were adjusted for potential confounders identified from the directed acyclic graph, including: maternal age (continuous), education level (none/primary/secondary/higher), household wealth quintile (poorest to richest), urban/rural residence, geopolitical region, religious affiliation, early marriage (marriage before age 18), and parity (number of live births). Under this framework, we define potential outcomes Y(a,m) as the value of the outcome that would be observed if the exposure were set to a and the mediator to m. Similarly, M(a) denotes the potential mediator value under exposure level a.

The natural direct effect (NDE) captures the effect of exposure on the outcome that would remain if the mediator were held at the value it would have taken in the absence of exposure:

The natural indirect effect (NIE) captures the effect that operates through changing the mediator:

The total effect decomposes as:

Identification of these effects requires four assumptions: (1) no unmeasured exposure-outcome confounding given C; (2) no unmeasured mediator-outcome confounding given A and C; (3) no unmeasured exposure-mediator confounding given C; and (4) no mediator-outcome confounder affected by exposure.³²

Estimation Approaches

We implemented three complementary estimation strategies.

Baron and Kenny Regression with Product-of-Coefficients

We fitted three logistic regression models: (1) the outcome model regressing Y on A alone to obtain the total effect c; (2) the mediator model regressing M on A to obtain path coefficient a; and (3) the full model regressing Y on A, M₁, M₂, M₃ simultaneously to obtain the direct effect c′ and mediator effects b₁, b₂, b₃. The indirect effect through each mediator was estimated as the product a_i × b_i. The proportion mediated was calculated as (c − c′)/c.

For the outcome model with binary outcome Y and binary mediators M:

For the mediator model:

Inverse Probability Weighting (IPW)

To address potential confounding of the mediator-outcome relationship, we estimated propensity scores for each mediator, conditional on the exposure and baseline covariates. The stabilized inverse probability weights were calculated as:

where the numerator is the marginal probability of the observed mediator given exposure and the denominator is the conditional probability given exposure and covariates. Weighted logistic regression models estimated the controlled direct effect of each mediator on the outcome.

Doubly Robust Estimation

We combined propensity score weighting with outcome regression using augmented inverse probability weighting (AIPW). This approach provides consistent estimates if either the propensity score model or the outcome model is correctly specified, offering protection against model misspecification.³³

Bootstrap Confidence Intervals

A complete case analysis was performed on 4,236 women with non-missing data on all variables (77.1% of fistula cases). Women with missing data were more likely to reside in rural areas (81.2% vs. 73.4%, p<0.001) and Northern regions (76.8% vs. 68.1%, p<0.001), but did not differ significantly by age, education, or treatment-seeking status. We conducted sensitivity analyses using multiple imputation by chained equations (MICE) with 50 imputed datasets (see Supplementary Methods 1 and Supplementary Table S1). Results were qualitatively identical to complete case analysis: geographic access barriers remained the only significant mediator (pooled OR=0.74; 95% CI: 0.61–0.90), while financial (pooled OR=1.09; 95% CI: 0.93–1.27) and autonomy barriers (pooled OR=0.97; 95% CI: 0.82–1.15) remained non-significant. Given the consistency of findings and the methodological challenges of imputing mediator values under missing-not-at-random mechanisms, we present complete case results in the main text with imputation results (See Supplementary Tables S1–Table S4 and Supplementary Information 1).

Additionally, we conducted sensitivity analyses using alternative mediator specifications, including: (1) count variables summing the number of barriers in each domain (range 0–3 for access barriers, 0–1 for financial barriers, 0–2 for autonomy barriers), and (2) examining individual barrier items separately rather than composites. Results were qualitatively similar across specifications, with geographic access barriers remaining the only significant mediator. We conducted sensitivity analyses using alternative mediator specifications, including count variables and individual barrier items (Supplementary Tables S2 and Table S3).

Given that indirect effects are products of coefficients and thus do not follow a normal distribution, we computed 95% confidence intervals using the percentile bootstrap method with 1,000 replications.³⁴ For each bootstrap sample, we re-estimated all model parameters and calculated indirect effects, yielding empirical distributions from which 2.5th and 97.5th percentiles were extracted.

Sensitivity Analysis

We conducted E-value sensitivity analyses to assess robustness to unmeasured confounding.³⁵ The E-value represents the minimum strength of association that an unmeasured confounder would need to have with both the exposure and outcome, conditional on measured covariates, to fully explain away the observed association. For an observed odds ratio (OR), the E-value is calculated as

For protective associations (OR < 1), we first converted to 1/OR before applying the formula. Higher E-values indicate greater robustness to unmeasured confounding.

All statistical analyses were conducted using R (version 4.3) and Python (version 3.11). The R packages included tidyverse for data manipulation, tableone for descriptive statistics, boot for bootstrap resampling, and WeightIt for inverse probability weighting. Python analyses utilized pandas, numpy, scipy, and statsmodels for regression modeling and mediation decomposition. Figures were generated using ggplot2 in R and Matplotlib in Python.

Results

Study Population

Of the 39,050 women of reproductive age surveyed in the Nigeria DHS 2024, 23,474 (60·1%) responded to the obstetric fistula symptom questions and were included in this analysis. Among the respondents, 5,496 women (23·4%) reported experiencing symptoms consistent with obstetric fistula (involuntary leakage of urine or feces following childbirth). Table 1 presents the characteristics of the study population stratified by fistula status. Women with fistula were less likely to have knowledge about fistula symptoms compared to those without fistula (71·0% vs 77·2%, p<0·001). Regarding healthcare barriers, women with fistula reported fewer financial barriers (46·4% vs 52·7%, p<0·001) and geographic access barriers (64·5% vs 68·9%, p<0·001) compared to women without fistula; however, this likely reflects selection effects related to healthcare engagement. There was no significant difference in high healthcare decision-making autonomy between groups (55·5% vs 56·3%, p=0·34). Results were robust to missing data handling and alternative specifications (Supplementary Tables S1, Table S3, and Table S4).

Table 1.

Characteristics of Study Population by Fistula Status

Characteristic	Overall (N=23,474)	Fistula (n=5,496)	No Fistula (n=17,978)	p-value
Knowledge
Has knowledge about fistula, n (%)	17,791 (75·8)	3,904 (71·0)	13,887 (77·2)	<0·001
Healthcare barriers
Money for treatment, n (%)	18,024 (49·3)	2,550 (46·4)	9,474 (52·7)	<0·001
Distance to facility, n (%)	24,233 (66·3)	3,545 (64·5)	12,388 (68·9)	<0·001
Transportation, n (%)	10,643 (29·1)	1,380 (25·1)	5,328 (29·6)	<0·001
Getting permission, n (%)	7,036 (19·2)	1,259 (22·9)	3,884 (21·6)	0·051
Not going alone, n (%)	2,964 (8·1)	407 (7·4)	1,350 (7·5)	0·860
Female provider concern, n (%)	7,314 (20·0)	1,039 (18·9)	4,082 (22·7)	<0·001
Drug availability concern, n (%)	5,022 (13·7)	632 (11·5)	2,841 (15·8)	<0·001
Provider availability concern, n (%)	1,717 (4·7)	324 (5·9)	809 (4·5)	<0·001
Number of barriers, mean (SD)	2·10 (1·38)	1·97 (1·38)	2·16 (1·37)	<0·001
Decision-making autonomy
Respondent decides alone, n (%)	7,289 (19·9)	891 (16·2)	3,651 (20·3)	<0·001
Joint with husband, n (%)	15,641 (42·8)	2,666 (48·5)	8,273 (46·0)	0·001
Husband only, n (%)	1,621 (4·4)	220 (4·0)	863 (4·8)	0·009
Other family member, n (%)	9,107 (24·9)	1,522 (27·7)	5,469 (30·4)	<0·001
High autonomy (self or joint), n (%)	20,316 (55·6)	3,051 (55·5)	10,122 (56·3)	0·340

Notes: Values are n (%) for categorical variables and mean ± SD for continuous variables. P-values from χ² tests for categorical variables and independent t-tests for continuous variables.

Obstetric Fistula Treatment Cascade

Table 2 illustrate the treatment cascade among women with obstetric fistula. It is important to note that the cascade reflects reported status at the time of survey interview and may not represent completed care trajectories. Women classified as “sought treatment but no surgery” may include those currently awaiting surgical appointments, those who were evaluated but deemed not suitable candidates for repair, those who were lost to follow-up during referral processes, and those who experienced misclassification of treatment type (eg., receiving medical management but not definitive surgical repair). Of the 5,496 women reporting fistula symptoms, only 3,572 (65·0%) sought treatment, representing a first treatment gap of 1,924 women (35·0%) who never engaged with healthcare services. Among those who sought treatment, only 180 (3·3% of all fistula cases; 5·0% of treatment-seekers) received surgical repair, representing a second treatment gap of 3,392 women (61·7%) who sought care but did not receive definitive surgical treatment. Ultimately, 380 (6·9%) women achieved fistula resolution, leaving 5,116 (93·1%) women with unresolved obstetric fistula. The largest attrition in the treatment cascade occurred between treatment-seeking and surgical repair, where 94·7% of women who sought care did not undergo surgery. This gap suggests a critical bottleneck in the fistula care continuum in Nigeria.

Table 2.

Obstetric Fistula Treatment Cascade

Stage	n	% of Total	Attrition from Previous Stage
Women with fistula symptoms	5,496	100·0	–
Sought treatment	3,572	65·0	1,924 (35·0%) did not seek treatment
Received surgical repair	180	3·3	3,392 (61·7%) sought care but no surgery
Fistula resolved	380	6·9	5,116 (93·1%) remain unresolved

Notes: Treatment cascade shows the proportion of women with fistula symptoms at each stage of care. Gaps represent attrition between consecutive stages. Percentages in parentheses are calculated relative to the total number of women with fistula symptoms (N=5,496).

Figure 1 shows the number of women at each stage of the treatment cascade, with treatment gaps annotated. Of 5,496 women with fistula symptoms, 65·0% sought treatment, 3·3% underwent surgical repair, and 6·9% achieved resolution. The largest attrition (61·7%) occurred between treatment-seeking and surgical receipt.

Figure 1.

Obstetric fistula treatment cascade among 5,496 women with fistula symptoms in Nigeria, 2024. The treatment cascade shows progression from fistula symptom onset through treatment-seeking and surgical receipt. Bold values in the figure indicate counts. Of 5,496 women reporting fistula symptoms, 38.3% (n=2,105) sought treatment. Among treatment-seekers, only 5.3% (n=112) received surgical repair, resulting in 61.7% attrition between treatment-seeking and surgery. Numbers and percentages are based on weighted sample estimates accounting for DHS complex survey design. Data source: Nigeria Demographic and Health Survey 2024.

Causal Mediation Analysis

Figure 2 presents the path diagram with estimated coefficients. Table 3 presents the regression results of the analysis.

Figure 2.

Directed acyclic graph depicting causal mediation pathways between fistula knowledge and treatment-seeking through three healthcare barrier mediators. Solid arrows indicate the natural direct effect (NDE) of knowledge on treatment-seeking; dashed arrows represent indirect pathways through mediators. Path coefficients (β) are presented on the log-odds scale from Baron-Kenny mediation models. Asterisk (*) indicates statistical significance at p < 0.05. Bold values in the figure indicate statistically significant associations. Fistula knowledge significantly increased perceptions of all three barrier types: geographic access barriers (β=0.59, p<0.05), financial barriers (β=0.50, p<0.05), and autonomy barriers (β=0.52, p<0.05). However, only geographic access barriers significantly reduced treatment-seeking (β=−0.32, p<0.01). All models adjusted for age, education, wealth quintile, urban/rural residence, region, religion, early marriage, and parity. The total effect (TE) of knowledge on treatment-seeking was not statistically significant (β=−0.13, 95% CI: −0.31 to 0.04), with approximately 15% proportion mediated through the examined pathways.

Abbreviations: CI, confidence interval; NDE, natural direct effect; NIE, natural indirect effect; TE, total effect.

Table 3.

Association Between Healthcare Barriers and Treatment-Seeking Among Women with Obstetric Fistula

Variable	Unadjusted OR (95% CI)	p-value	Adjusted OR (95% CI)	p-value	IPW-Adjusted OR (95% CI)	p-value
Fistula knowledge	0·88 (0·73–1·05)	0·14	0·89 (0·74–1·07)	0·23	–	–
Financial barriers	1·05 (0·90–1·23)	0·53	1·11 (0·95–1·31)	0·19	1·07 (0·91–1·25)	0·42
Access barriers	0·73 (0·60–0·89)	0·002	0·73 (0·59–0·88)	0·002	0·76 (0·62–0·92)	0·004
Autonomy barriers	0·97 (0·82–1·16)	0·77	0·99 (0·84–1·18)	0·94	0·95 (0·80–1·13)	0·57

Notes: All models adjusted for age, education, wealth quintile, urban/rural residence, region, religion, early marriage, and parity. P-values less than 0.05 indicate statistical significance. Adjusted model includes all barriers, fistula knowledge and also controlled for the following baseline covariates: age (continuous), education (none/primary/secondary/higher), household wealth quintile (poorest to richest), urban/rural residence, geopolitical region, religious affiliation, early marriage (<18 years), and parity. IPW = Inverse Probability Weighting, adjusted for fistula knowledge and baseline covariates.

Abbreviations: OR, odds ratio; CI, confidence interval; IPW, inverse probability weighting.

Total and Direct Effects

In the unadjusted model examining the relationship between fistula knowledge and treatment-seeking among women with fistula (n=4,236 with complete data), knowledge was not significantly associated with treatment-seeking (OR=0·88; 95% CI: 0·73–1·05; p=0·14). After adjusting for healthcare barriers (financial, access, and autonomy), the direct effect of knowledge remained non-significant (OR=0·89; 95% CI: 0·74–1·07; p=0·23).

Mediator Effects on Treatment-Seeking

Geographic access barriers were significantly associated with reduced treatment-seeking (OR=0·73; 95% CI: 0·59–0·88; p=0·002), indicating that women reporting distance or transportation barriers had 27% lower odds of seeking treatment for fistula (Table 3 and Figure 2). Financial barriers showed no significant association with treatment-seeking (OR=1·11; 95% CI: 0·95–1·31; p=0·19), nor did autonomy barriers (OR=0·99; 95% CI: 0·84–1·18; p=0·94). Figure 3 displays barrier prevalence by fistula status, and Figure 4 presents the forest plot of barrier effects with unadjusted and IPW-adjusted estimates. The lack of statistically significant associations for financial and autonomy barriers should not be interpreted as evidence of unimportance. Wide confidence intervals suggest imprecision rather than definitive null effects, and these barriers may operate at different stages of the cascade (eg., affecting surgical receipt rather than initial treatment-seeking) or may be highly correlated with geographic access such that their independent effects are difficult to isolate.

Figure 3.

Healthcare barrier prevalence by fistula status, Nigeria DHS 2024. Grouped bar chart comparing the prevalence of eight healthcare barriers between women with and without fistula symptoms. Error bars represent 95% confidence intervals. Geographic access barriers (distance to facility, transportation) were the most commonly reported barriers in both group.

Figure 4.

Healthcare barrier effects on treatment-seeking among women with obstetric fistula. Forest plot displaying odds ratios and 95% confidence intervals for the association between each barrier type and treatment-seeking behavior among women with fistula. Both unadjusted estimates and inverse probability-weighted (IPW) estimates are shown. Vertical dashed line indicates OR = 1.0 (no association). Only geographic access barriers showed a statistically significant association (OR=0.73; 95% CI: 0.59–0.88, p < 0.05).

Figure 3 (Grouped bar chart) compares the prevalence of eight healthcare barriers between women with and without fistula. Geographic access barriers (distance, transportation) were the most commonly reported barriers in both groups.

Mediation Decomposition

Table 4 presents the results of the causal mediation analysis using the difference method consistent with the counterfactual framework. We estimated a total effect of β=−0·13 (95% CI: −0·31 to 0·04), a natural direct effect of β=−0·11 (95% CI: −0·29 to 0·06), and a total natural indirect effect of β=−0·02 (95% CI: −0·05 to 0·01). Approximately 16·6% of the total effect was mediated through healthcare barriers. Pathway-specific analyses using the product-of-coefficients method with 1,000 bootstrap replications revealed that the indirect effect through access barriers was statistically significant (β=−0·19; 95% CI: −0·33 to −0·07; p=0·002), whereas indirect effects through financial barriers (β=0·05; 95% CI: −0·02 to 0·14; p=0·19) and autonomy barriers (β=−0·003; 95% CI: −0·10 to 0·08; p=0·94) were not significant.

Table 4.

Causal Mediation Analysis: Decomposition of Effects

Effect	Estimate (β)	Bootstrap SE	95% CI	p-value
Total effect	−0·134	0·091	(−0·31, 0·04)	0·14
Natural Direct Effect	−0·112	0·093	(−0·29, 0·06)	0·23
Total Natural Indirect Effect	−0·022	0·016	(−0·05, 0·01)	0·18
Pathway-specific indirect effects
Via Financial Barriers	0·054	0·043	(−0·02, 0·14)	0·19
Via Access Barriers	−0·190	0·066	(−0·33, −0·07)	0·002
Via Autonomy Barriers	−0·003	0·047	(−0·10, 0·08)	0·94
Proportion Mediated	16·6%	–	–	–

Notes: β = regression coefficient on log-odds scale. Bootstrap confidence intervals based on 1,000 replications. All models adjusted for maternal age, education level, household wealth quintile, urban/rural residence, geopolitical region, religious affiliation, early marriage (marriage before age 18), and parity. P-values less than 0.05 indicate statistical significance.

Abbreviations: CI, confidence interval; NDE, natural direct effect; NIE, natural indirect effect.

Inverse Probability Weighting Analysis

To address potential confounding, we estimated propensity score-weighted effects of barriers on treatment-seeking. After IPW adjustment, geographic access barriers remained significantly associated with reduced treatment-seeking (OR=0·76; 95% CI: 0·62–0·92), consistent with the unadjusted estimate (Table 3 and Figure 4). Financial barriers (IPW-adjusted OR=1·07; 95% CI: 0·91–1·25) and autonomy barriers (IPW-adjusted OR=0·95; 95% CI: 0·80–1·13) remained non-significant after weighting. Doubly robust estimation using augmented inverse probability weighting confirmed these results. Women with access barriers had a 3·9 percentage point lower probability of seeking treatment (risk difference: −0·039) compared to those without access barriers, corresponding to a risk ratio of 0·95. Figure 4 is a Forest plot displaying odds ratios and 95% confidence intervals for the association between each barrier type and treatment-seeking, with both unadjusted and inverse probability-weighted estimates. Only access barriers showed a statistically significant association (OR=0·73; 95% CI: 0·59–0·88).

Sensitivity Analysis

Table 5 presents the E-value sensitivity analysis for unmeasured confounding. For the association between access barriers and treatment-seeking (OR=0·73), the E-value was 2·08, indicating that an unmeasured confounder would need to be associated with both access barriers and treatment-seeking by a risk ratio of at least 2·08 to fully explain away the observed association. The E-value for the confidence interval limit was 1·50, suggesting moderate robustness to unmeasured confounding. For financial barriers (E-value=1·49) and autonomy barriers (E-value=1·19), the lower E-values and confidence intervals crossing the null indicate that these associations are less robust and could be explained by relatively weak unmeasured confounders. E-value represents the minimum strength of association an unmeasured confounder would need with both the exposure and outcome to fully explain away the observed association, conditional on measured covariates.

Table 5.

E-Value Sensitivity Analysis for Unmeasured Confounding

Association	OR	95% CI	E-value (Point)	E-value (CI)	Interpretation
Access barriers → Treatment-seeking	0·73	(0·59, 0·88)	2·08	1·50	Moderate robustness
Financial barriers → Treatment-seeking	1·11	(0·95, 1·31)	1·49	1·00	Low robustness (CI crosses 1)
Autonomy barriers → Treatment-seeking	0·99	(0·84, 1·18)	1·19	1·00	Low robustness (CI crosses 1)

Notes: E-value = minimum strength of association an unmeasured confounder would need with both exposure and outcome to fully explain away the observed association, conditional on measured covariates. Higher E-values indicate greater robustness to unmeasured confounding. E-values calculated using the formula: E-value = OR + √[OR × (OR − 1)] for protective associations (OR < 1), first converting to 1/OR. → indicates direction of association (barrier → treatment-seeking). All models controlled for the following baseline covariates: age (continuous), education (none/primary/secondary/higher), household wealth quintile (poorest to richest), urban/rural residence, geopolitical region, religious affiliation, early marriage (<18 years), and parity.

Geographic Heterogeneity

Figure 5 shows the geographic distribution of fistula prevalence across 1,380 survey clusters. Cluster-level fistula prevalence ranged from 0% to 93·3%, with a mean of 14·8%. A total of 134 clusters (9·7%) had prevalence exceeding 40%, while 423 clusters (30·7%) had zero reported cases. Cluster-level fistula prevalence was positively correlated with knowledge rates (r=0·31, p<0·001), suggesting that awareness may be higher in communities with greater fistula burden, potentially reflecting lived experience or targeted health education efforts.

Figure 5.

Geographic variation in obstetric fistula prevalence (N = 1,380 clusters). (A) Histogram showing the distribution of cluster-level fistula prevalence across 1,380 survey clusters (mean=14.8%, range=0–93.3%). Vertical dashed line indicates mean prevalence. (B) Scatterplot of cluster fistula prevalence versus knowledge rate, with points colored by the mean number of barriers reported. Fitted regression line shows positive correlation (r=0.31, p<0.001), suggesting that awareness may be higher in communities with greater fistula burden.

From Figure 5, (A) Histogram shows the distribution of cluster-level fistula prevalence across 1,380 survey clusters (mean=14·8%, range=0–93·3%). (B) Scatterplot of cluster fistula prevalence versus knowledge rate, colored by the mean number of barriers, with a fitted regression line (r=0·31, p<0·001).

Discussion

Our findings should be interpreted with appropriate caution regarding causal inference. While we employed the counterfactual framework for mediation analysis with explicit identification assumptions, the cross-sectional design precludes definitive establishment of temporal ordering. Healthcare barriers were assessed at the time of survey interview and may not reflect barriers present when treatment-seeking decisions were made. The associations we report are consistent with causal mediation under the stated assumptions (no unmeasured confounding of exposure-outcome, mediator-outcome, and exposure-mediator relationships, and no mediator-outcome confounders affected by exposure), but alternative explanations including reverse causation cannot be ruled out. For instance, women who successfully sought treatment may retrospectively minimize perceived barriers, while those who did not seek care may emphasize barriers to justify their decisions. Despite these limitations, the magnitude, specificity, and consistency of findings across multiple analytic approaches support substantive interpretation.

This study provides the first comprehensive causal mediation analysis of the obstetric fistula treatment cascade at a national scale, revealing several findings with important implications for maternal health policy. Among 5,496 Nigerian women with fistula symptoms, we documented a treatment cascade characterized by profound failure: while 65% sought treatment, only 3.3% underwent surgical repair, and 6.9% achieved resolution, leaving 93.1% with unresolved fistula. Using rigorous causal inference methodology, we identified geographic access barriers, distance to facilities, and transportation as the sole statistically significant mediating pathways associated treatment-seeking, with women reporting these barriers having 27% lower odds of engaging with care.

The cascade must be interpreted with recognition of its cross-sectional nature. The 61.7% gap between treatment-seeking and surgical receipt likely reflects a heterogeneous group: women awaiting scheduled surgeries, those who sought care but were referred to distant facilities they could not reach, those evaluated and placed on waitlists at overburdened surgical centers, those who received medical rather than surgical management, and those whose fistulas resolved spontaneously. Without longitudinal data tracking treatment trajectories, we cannot definitively partition these categories.^8,36 Nevertheless, the magnitude of this gap with 94.7% of treatment-seekers not receiving surgery, far exceeds plausible proportions for spontaneous resolution or medical management, indicating substantial system-level bottlenecks.^8,37 This suggests that the primary bottleneck lies within the health system, such as insufficient surgical capacity, referral system failures, or barriers encountered after the initial presentation, rather than in women’s motivation to seek care.^38,39 This finding aligns with observations from Ethiopia and Uganda, which documented substantial delays between the first healthcare contact and definitive surgical repair, often spanning years.^40,41 At current treatment rates, and assuming no new cases, it would take over 300 years to clear Nigeria’s backlog of untreated fistula cases.⁴²

The identification of geographic access barriers as significantly associated with treatment-seeking, while financial and autonomy barriers were not, challenges assumptions about which obstacles are most consequential at different cascade stages. The finding that fistula knowledge significantly increased perceptions of all barrier types (β = 0.50–0.59) warrants careful interpretation. Rather than indicating that knowledge creates barriers, this likely reflects that informed women are more aware of and able to articulate the structural constraints they face.^43,44 Women who have heard of fistula and its treatment may better understand the geographic distance to specialized centers, the financial costs of surgery and lodging, and the autonomy required to pursue extended treatment. This suggests that awareness campaigns, while valuable for symptom recognition, do not reduce objective barriers and may instead reveal the magnitude of system-level constraints.^45,46

In our adjusted models, financial barriers were not independently associated with treatment-seeking (OR=1.11; 95% CI: 0.95–1.31). This finding requires nuanced interpretation and should not be read as evidence that financial constraints are unimportant. Several alternative explanations merit consideration. First, financial barriers may exert their primary influence at later stages of the cascade specifically, in determining whether women who seek treatment actually undergo surgery rather than in the initial decision to seek care. Second, financial and geographic barriers may be highly collinear; women facing prohibitive transportation costs or distant facilities may frame their constraint as geographic rather than financial when both are operative.⁴⁷ Third, the single-item measure (“getting money for treatment is a big problem”) may not adequately capture the multidimensional nature of economic barriers, including indirect costs (lodging, meals, lost income during recovery) and catastrophic expenditure relative to household resources. Finally, Nigeria’s history of subsidized fistula repair campaigns may have reduced financial barriers specifically for treatment-seeking, even if costs remain prohibitive at the surgical stage.^48,49

Our findings support the growing evidence for decentralized fistula care delivery models. Mobile surgical outreach programs, wherein specialist teams travel to regional or district hospitals to perform repairs, have demonstrated high success rates and improved access in Tanzania, Malawi and Ethiopia.^37,50 Transportation voucher programs, successfully implemented for facility-based deliveries in multiple African countries, could be adapted to ort fistula patients’ travel to surgical sites.⁵¹ Task-shifting approaches that train general surgeons or medical officers in basic fistula repair could expand the capacity at facilities closer to where women live, although quality assurance mechanisms are essential.⁵²

An E-value of 2.08 for the geographic access barrier association indicates moderate robustness to unmeasured confounding. An unmeasured confounder would need to be associated with both access barriers and treatment-seeking by a risk ratio of at least 2.08 to fully explain the observed association. Potential unmeasured confounders include socioeconomic status (beyond what is captured by the financial barrier measure), education level, fistula severity, urban/rural residence, and regional variations in healthcare infrastructure. Although we cannot rule out residual confounding, the magnitude of the E-value and the biological plausibility of geographic barriers associated with treatment engagement support a causal interpretation.

Several methodological strengths enhance the confidence in our findings. The use of nationally representative DHS data with standardized obstetric fistula questions provides an unprecedented sample size and generalizability. Our application of the counterfactual framework for causal mediation, with explicit articulation of identification assumptions, represents a substantial advancement over previous descriptive studies on obstetric fistula barriers. The triangulation of results across multiple estimation approaches, Baron and Kenny regression, IPW, and doubly robust estimation with consistent findings, strengthens inferential validity.^25,33 Bootstrap confidence intervals appropriately account for the non-normal distribution of indirect effects.

This study has some important limitations. First, the cross-sectional design precludes establishment of temporal ordering; healthcare barriers were assessed at the time of interview and may not reflect barriers present at the time of treatment-seeking decisions. Second, fistula status was based on self-reported symptoms rather than clinical examination, potentially introducing misclassification errors. However, DHS fistula questions have been validated against clinical diagnoses in multiple settings, with acceptable sensitivity and specificity.⁵³ Our operationalization of mediators has several limitations. The composite barrier variables were dichotomized (any vs. none), which may dilute dose-response relationships and underestimate effect sizes. Financial barriers were assessed through a single item (“getting money for treatment is a big problem”), which may not capture the multidimensional nature of economic constraints including catastrophic costs of surgery, lodging during recovery, transportation expenses, and lost income. Autonomy barriers similarly relied on two items (permission to go for treatment, not wanting to go alone) and did not directly assess spousal refusal, stigma, or community-level constraints. Future research with more detailed barrier assessments may reveal stronger or more nuanced mediating pathways. Third, the mediator-outcome relationship may be confounded by factors that we could not measure, including healthcare quality, social support, and community-level stigma. Fourth, complete case analysis (77% of fistula cases) may introduce selection bias if missingness is related to both barriers and treatment-seeking, although sensitivity analyses using multiple imputation yielded similar results. Alternative mediator specifications yielded similar findings (Supplementary Table S2). Fifth, we could not distinguish between different types of fistula (vesicovaginal vs. rectovaginal) or assess fistula severity, which may modify barrier effects.

The findings have direct implications for Nigeria’s National Strategic Framework for Fistula Elimination and global efforts to achieve SDG target of ending preventable maternal morbidity by 2030. Current strategies emphasizing awareness generation and subsidized surgery should be complemented by substantial investments to reduce geographic barriers. Specific recommendations include the following: (1) expansion of mobile surgical outreach programs to bring fistula repair services to district hospitals in high-burden regions; (2) implementation of transportation voucher schemes specifically targeting fistula patients; (3) geographic mapping of fistula cases and surgical capacity to optimize service placement; and (4) strengthening referral systems to ensure that women who present to primary facilities are effectively linked to surgical services.

Conclusion

This causal mediation analysis of Nigerian women with obstetric fistula showed that geographic access barriers and distance suggest the critical pathways associated with treatment, with women reporting these barriers having 27% lower odds of seeking care. The dominance of geographic access barriers as the sole significant mediator reflects measurable system-level deficiencies: fewer than 100 fistula surgeons for a population exceeding 200 million, concentration of repair services in approximately 20 urban centers, and poorly defined referral pathways between primary care and specialized facilities. These structural features of the Nigerian health system create a spatial bottleneck that neither knowledge nor individual motivation can overcome. Contrary to assumptions about financial constraints and awareness as key obstacles, our findings revealed that neither financial nor autonomy-related barriers significantly mediated treatment-seeking after accounting for geographic accessibility. With 93% of affected women having unresolved fistula, and our finding that the primary bottleneck occurs between treatment-seeking and surgical receipt rather than symptom onset and care engagement, women are willing to seek care, but the system fails to deliver surgical services. The finding that fistula knowledge alone was insufficient to drive treatment-seeking reinforces the argument that awareness campaigns, while necessary for symptom recognition, cannot overcome structural barriers to care. Knowledge appears to be a necessary but not sufficient condition: it must be accompanied by accessible, affordable, and geographically proximate services to translate into care engagement. This evidence supports a shift in fistula elimination strategies from awareness campaigns to supply side interventions that address the geographic maldistribution of surgical capacity. Achieving the Sustainable Development Goal of ending preventable maternal morbidity by 2030 requires restructuring fistula care through mobile surgical outreach, transportation vouchers, decentralization of repairs, and strategic mapping to optimize surgical placement. Without addressing distance-related barriers, the treatment gap of over five thousand Nigerian women living with preventable disabilities will persist for generations.

Funding Statement

The funders had no role in study design, data collection, analysis, interpretation, or manuscript preparation. The corresponding author had full access to all data and was responsible for the decision to submit the manuscript for publication.

Data Sharing Statement

The Nigeria DHS 2024 dataset is publicly available from The DHS Program (https://dhsprogram.com) upon registration and approval of a data use agreement.

Ethical Considerations

The 2024 NDHS protocol received ethical approval from the National Health Research Ethics Committee of Nigeria (Approval Number: NHREC/01/01/2007-10/12/2023) and the ICF Institutional Review Board (FWA00000845). All participants provided written informed consent prior to interview. Geographic coordinates were displaced 2–5 km (urban) or 5–10 km (rural) to protect respondent confidentiality. This secondary analysis of de-identified publicly available data was determined to be exempt from additional institutional review, as it involves analysis of existing de-identified data that cannot be linked to individual subjects.

Author Contributions

All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

All authors declare no conflicts of interest in this work.