We conducted a hospital-based cross-sectional study with prospective data collection and an analytical component at the obstetrics and gynecology emergency department of the Centre Hospitalier Universitaire Départemental du Borgou et de l’Alibori (CHUD-B/A), northern Benin. The study was conducted over a six-month period, from 1 January to 30 June 2022. The study population consisted of women presenting to the obstetrics and gynecology emergency department during the study period who had been referred or transferred from another health facility for obstetric emergency care. The CHUD-B/A is a referral hospital providing specialized obstetric and gynecological care and receiving patients referred from health facilities within the surrounding health zones.

All women referred or transferred to the obstetrics and gynecology emergency department of CHUD-B/A during the study period were eligible for inclusion.

Women were included if they:

had been referred or transferred from another health facility; were admitted to the obstetrics and gynecology emergency department between 1 January and 30 June 2022; had provided consent to participate in the study.

Women whose records could not be completed because they were lost to follow-up or left the hospital before the necessary information could be collected were excluded from the final analysis.

A census sampling approach was used. All eligible women referred or transferred to CHUD-B/A during the study period were consecutively enrolled. No a priori sample-size calculation was therefore performed, as the study aimed to include the entire accessible population during the predefined six-month period.

The primary outcomes were:

referral dysfunction, defined as failure to meet the predefined quality criteria of the obstetric referral process; and counter-referral dysfunction, defined as failure to meet the predefined quality criteria of the counter-referral process.

The study also collected sociodemographic, geographical, organizational, clinical and process-related variables.

These included maternal age, occupation, area of residence, type of referring facility, distance between the referring facility and CHUD-B/A, consent to referral, availability and completeness of the referral form, venous access, pre-referral treatment, notification of the receiving hospital, means of transport, health-worker accompaniment, arrangements for reception, and time elapsed before initial management.

For counter-referral, variables included transmission of feedback to the referring facility, availability and completion of the counter-referral documentation, information on treatment provided and recommendations for subsequent management.

Definition and assessment of referral dysfunction:The assessment of referral quality was based on the national standards and procedures for referral and counter-referral applicable in Benin. The referral process was assessed using predefined criteria covering the main stages of the referral pathway: consent to referral; availability of a referral form; completeness of the referral form; pre-referral counselling; establishment of venous access; pre-referral stabilization; communication with the receiving facility; appropriate transport; accompaniment by a health professional when indicated; treatment during transfer; and preparation for reception and initial management at the referral hospital. For the purpose of this study, a criterion was considered adequately fulfilled when the predefined threshold of 70% was achieved. Failure to achieve this threshold was considered indicative of dysfunction for the corresponding referral component.

Definition and assessment of counter-referral dysfunction:The counter-referral process was assessed using the national referral and counter-referral standards. The following criteria were considered: availability of a counter-referral form; adequate completion of the counter-referral form; documentation of treatment provided; transmission of recommendations for continued management; and transmission of feedback to the referring facility within the expected timeframe. As for the referral process, a compliance threshold of 70% was used to identify dysfunction. Because the study evaluated several individual components of the referral pathway, the results are presented both for each criterion and for the overall assessment of referral and counter-referral dysfunction.

Data were collected prospectively using a structured data-collection questionnaire developed from the national referral and counter-referral standards. Information was obtained through interviews with the referred women or, when necessary, their accompanying relatives. The collected information was subsequently cross-checked and completed using the patients’ medical records, referral documents, hospital registers and counter-referral records.

The data-collection process was supervised to improve completeness and consistency. Where information was unavailable, the corresponding variable was recorded as missing rather than inferred.

Data were entered using EpiData version 3.1 and subsequently exported for analysis using Epi Info version 7.2 and Stata SE version 13.0.

Records with insufficient information to assess the main study outcomes were excluded from the final analytical population. For individual variables, missing information was retained as missing and was not replaced by an assumed value. The number of observations contributing to each analysis was reported when it differed from the total study population.

Categorical variables were summarized using frequencies and percentages. Quantitative variables were described using means and standard deviations when approximately normally distributed and, where appropriate, medians and interquartile ranges.

The prevalence of each referral and counter-referral quality indicator was calculated using the number of women for whom the relevant information was available as the denominator.

Bivariate analyses were performed to explore associations between potential explanatory variables and referral or counter-referral dysfunction. The Pearson χ² test or Fisher’s exact test, as appropriate, was used for categorical variables.

Variables with a p value < 0.20 in the bivariate analysis were considered for inclusion in the multivariable model. Multivariable analysis was then performed to identify factors independently associated with referral and counter-referral dysfunction.

Because the outcomes were common in this study, associations were expressed as adjusted prevalence ratios (aPRs) with 95% confidence intervals (CIs) rather than odds ratios. Statistical significance was defined as a two-sided p value < 0.05.

The study was conducted in accordance with the principles of the Declaration of Helsinki and applicable national requirements for research involving human participants. Administrative authorization was obtained from the CHUD-B/A and the relevant hospital authorities. Written informed consent was obtained from eligible participants before enrolment. Confidentiality was maintained throughout the study. Data were collected and analyzed anonymously and were used exclusively for research purposes, in accordance with Law No. 2009-09 of 22 May 2009 on the protection of personal data in the Republic of Benin. The research proposal received approval from the administrative authorities of the Ordre de Malte hospital and from the local ethics committee of the University of Parakou, under reference: 0404/CLERB-UP/P/SP/R/SA.

Between 1 January and 30 June 2022, 1735 women were admitted to the obstetrics and gynecology emergency department of CHUD-B/A. Of these, 919 (52.9%) were referred or transferred from another health facility. Six records were excluded because of incomplete information or incomplete follow-up, leaving 913 women in the final analytical population (Figure 1).

Figure 1. Flow diagram of women admitted, referred/transferred, excluded and included in the final analysis at CHUD-B/A, January-June 2022.

The mean age of the participants was 26.8 ± 6.6 years (range, 14 - 57 years). Women aged 20 - 29 years constituted the largest age group (483/913; 52.9%). Housewives were the most represented occupational group (405/913; 44.4%). Most women were referred from public health facilities (785/913; 86.0%) and from urban areas (580/913; 63.5%). The referring facility was located more than 10 km from CHUD-B/A for 336 women (36.8%).

Overall, 838 women (91.8%) had consented to referral. A referral form was available for 746 women (81.7%), while a standard Ministry of Health referral form was available for 404 (44.2%). Among women for whom a referral form was available, 633 (84.9%) had adequately completed documentation. The sociodemographic and referring-facility characteristics of the study population are presented in Table 1.

Table 1. Sociodemographic characteristics and characteristics of referring facilities among women referred to CHUD-B/A, 2022 (N = 913).

Among the 913 referred women, 654 (71.6%) had venous access established before transfer. Only 63 (6.9%) had documented emergency treatment or stabilization during the pre-referral period or transfer. Most women were transferred using non-medicalized transport (813/913; 89.1%), while 100 (11.0%) benefited from medicalized transport. Motorcycles were the most frequently used means of transport (642/913; 70.3%), followed by ambulances (93/913; 10.2%) and taxis or tricycles (88/913; 9.6%). Only 13 women (1.4%) were accompanied by a health professional during transfer. Prior notification of CHUD-B/A was documented for only 161 women (17.6%), whereas 752 (82.4%) arrived without prior notification. Specific arrangements for receiving the patient had been made in 141 cases (15.4%). Among the 913 women, 539 (59.0%) received initial care within 10 minutes of arrival, while 374 (41.0%) waited at least 10 minutes before receiving initial care. The mean duration of stay at the referring facility before transfer was 1.82 ± 2.88 hours among women transferred within 24 hours, with a median of 1 hour. Among women transferred after 24 hours, the mean duration was 2.04 ± 2.23 days, with a median of 1 day. The distribution of the main conditions and quality indicators of the referral process is presented in Table 2.

Table 2. Conditions of transfer and compliance with predefined quality criteria for the referral process among women referred to CHUD-B/A, 2022 (N = 913).

*Denominator: women for whom a referral form was available (n = 746).

The lowest levels of compliance concerned health-professional accompaniment (1.4%), documented treatment during transfer (6.9%), prior notification of the receiving hospital (17.6%), medicalized transport (11.0%), and arrangements for reception at the referral hospital (15.4%). In contrast, compliance was higher for consent to referral (91.8%), availability of a referral form (81.7%), adequate completion of the available referral form (84.9%), and establishment of venous access (71.6%) (Figure 2).

Figure 2. Adherence to key components of the obstetric referral system (N = 913).

Table 3. Maternal outcomes, duration of hospitalization and counter-referral indicators among women referred to CHUD-B/A, 2022 (N = 913).

At the end of hospitalization, 820 women (89.8%) were discharged alive after clinical stabilization. Forty women (4.4%) left the hospital against medical advice, 16 (1.8%) absconded before formal discharge, and 37 (4.1%) died. Thus, the four recorded maternal outcomes accounted for the entire study population. The median duration of hospitalization was 4 days. Overall, 747 women (81.8%) were hospitalized for less than 7 days, while 103 (11.3%) remained hospitalized for 7 days or longer. Duration of hospitalization was not documented in 63 cases (6.9%). Among women with documented maternal complications, the most frequently reported complications were sepsis (43.2%), hemorrhagic complications (24.3%), and anemia (16.2%). The counter-referral process showed substantial deficiencies. Documented feedback to the referring facility was available for only 323 women (35.4%). Treatment information was documented for 298 women (32.6%), while a standard counter-referral form was available for 132 (14.5%). Specific recommendations for continued management at the referring facility were documented in only 38 women (4.2%). Maternal outcomes, duration of hospitalization and counter-referral indicators are presented in Table 3.

In multivariable analysis, several characteristics were independently associated with referral dysfunction. Maternal age was significantly associated with referral dysfunction (p = 0.035), with a higher prevalence of dysfunction observed among younger women. The type of referring health facility was also associated with referral dysfunction (p = 0.006). Women referred from private facilities had a higher prevalence of referral dysfunction than those referred from public facilities (adjusted prevalence ratio [aPR] 1.14, 95% CI 1.06 - 1.24). Distance between the referring facility and CHUD-B/A was significantly associated with referral dysfunction (p = 0.018), with increasing distance associated with a higher prevalence of dysfunction. Rural origin was independently associated with referral dysfunction (p = 0.033); women originating from rural areas had a higher prevalence of referral dysfunction than those from urban areas (aPR 1.08, 95% CI 1.01 - 1.16). The time elapsed before initial management was also associated with referral dysfunction (p = 0.004). Women who waited at least 10 minutes before receiving initial care had a higher prevalence of referral dysfunction (aPR 1.11, 95% CI 1.04 - 1.19).

Distance between the referring facility and CHUD-B/A was strongly associated with counter-referral dysfunction (p < 0.001), with longer distances associated with a higher prevalence of dysfunction. Rural origin was also associated with counter-referral dysfunction (p = 0.002). Women from rural areas had a higher prevalence of counter-referral dysfunction than those from urban areas (aPR 1.21, 95% CI 1.07 - 1.36). Maternal outcome was significantly associated with counter-referral dysfunction (p < 0.001). Counter-referral dysfunction was more frequent among women who died during hospitalization (aPR 1.79, 95% CI 1.62 - 1.98). The complete multivariable analysis of factors associated with referral and counter-referral dysfunction is presented in Table 4.

Table 4. Factors independently associated with referral and counter-referral dysfunction among women referred to CHUD-B/A, 2022.

In this study, referred or transferred women represented 52.9% of all women admitted to the obstetric and gynecological emergency department. This proportion is comparable to the 51.5% reported by Tchaou et al. among obstetric emergencies managed at the University Hospital of Parakou [8]. It is higher than proportions reported in several studies from West Africa, including those conducted in Mali, Senegal and Burkina Faso [9]-[11]. These differences may reflect variations in the organization of referral networks, the geographical distribution of health facilities, accessibility to emergency obstetric care and the criteria used to define a referral or evacuation. They may also reflect differences in the level of the receiving facility and in the populations served. Nevertheless, the high proportion of referred women observed in the present study underlines the central role of CHUD-B/A within the regional obstetric care network. A high referral burden is not necessarily an indicator of poor performance. A functional referral system should enable women with complications that cannot be managed at peripheral facilities to reach an appropriate level of care. The critical issue is therefore not simply the proportion of women referred, but whether referral occurs in a timely, coordinated and clinically appropriate manner. This distinction is particularly important in obstetric emergencies, where delays may rapidly affect maternal and fetal outcomes [6][7].

The findings indicate substantial heterogeneity in the quality of the referral pathway. Consent to referral was obtained in 91.8% of cases and a referral form was available for 81.7%. These findings suggest that the formal decision to refer and basic referral documentation were relatively well established. However, other components of the referral process were considerably weaker. Only 17.6% of women had been announced to the receiving hospital before transfer, 15.4% had documented arrangements for reception, and only 1.4% were accompanied by a health professional. Medicalized transportation was also uncommon. These findings are consistent with the difficulties described in the national referral and counter-referral guidelines, which emphasize communication between facilities, appropriate transportation, pre-referral management and preparation of the receiving facility as essential components of an effective referral system [5]. The problem therefore appears less related to the formal decision to refer than to the operational coordination of the transfer.

The predominance of non-medicalized transportation is particularly important. In our study, almost nine out of ten women were transferred using non-medicalized transport, and motorcycles were the most common means of transportation. This contrasts with studies from settings where ambulance-based evacuation is more developed [13][18]. For example, Traoré reported that more than half of obstetric referrals or evacuations in Mali were performed by ambulance, whereas health-worker accompaniment remained limited [2]. Other studies have also demonstrated substantial variability in the means of transportation used for obstetric evacuation [9][10]. The predominance of motorcycles in the present study may partly reflect the geographical and socioeconomic context of northern Benin, where motorcycles are widely accessible and may be faster or more readily available than formal ambulance services. However, they provide limited capacity for monitoring, stabilization and emergency treatment during transfer. Thus, the accessibility advantage of motorcycles does not necessarily translate into a safe obstetric transfer.

One of the most important findings was the low proportion of women for whom CHUD-B/A had been notified before transfer. Only 17.6% of women were announced to the receiving facility, and arrangements for reception were documented in only 15.4%. This represents a critical weakness because referral is a process involving at least two facilities rather than a simple physical movement of a patient. Advance communication allows the receiving team to anticipate the patient’s arrival, mobilize personnel and equipment and prepare for emergency interventions. The absence of such communication may contribute to delays after arrival, particularly when several emergency patients are simultaneously managed.

In the present study, 41.0% of women waited at least 10 minutes before receiving initial care. Although 10 minutes is a relatively short interval in comparison with total hospitalization time, it may be clinically relevant in women presenting with severe hemorrhage, eclampsia, sepsis or other rapidly progressive obstetric emergencies. The observed association between a waiting time of at least 10 minutes and referral dysfunction supports the importance of considering the entire referral pathway rather than focusing exclusively on the journey from the referring facility to the hospital. This finding is consistent with the third component of the three-delays framework, which concerns delays in receiving appropriate care after arrival at a health facility [6].

Distance from the referring facility to CHUD-B/A was independently associated with referral dysfunction. Rural origin was also associated with both referral and counter-referral dysfunction. These findings are biologically and organizationally plausible. Greater distance increases travel time and exposes patients to a greater number of potential barriers, including poor road conditions, limited transportation availability and communication difficulties. The geographical dimension of access is particularly relevant in rural settings, where health facilities may have fewer qualified staff, fewer emergency resources and weaker transport links. Previous work has similarly demonstrated the importance of distance and accessibility in maternal healthcare utilization and referral outcomes [1][7][15][17]. The three-delays framework also identifies the ability to reach an appropriate health facility as a major determinant of maternal outcome [6]. Importantly, distance itself cannot easily be modified in the short term. However, its consequences can be reduced through better organization of referral networks, decentralized emergency stabilization, reliable communication systems, strategically positioned ambulances, improved road access and clear protocols for rapid transfer. The national referral guidelines provide an organizational framework for these interventions [5]. The association between rural origin and counter-referral dysfunction is also noteworthy. A referral system may function reasonably well for transferring a woman towards a higher-level facility while failing to maintain communication with the original facility after discharge. This suggests that geographical barriers affect not only emergency transfer but also continuity of care.

Counter-referral was the weakest component of the system assessed in this study. Only 35.4% of women had documented feedback to the referring facility, treatment information was documented in 32.6%, and recommendations for continued care were recorded in only 4.2%. These findings suggest that the referral pathway is largely unidirectional: peripheral facilities send patients to the referral hospital, but information does not consistently return to the referring facility. Such a system compromises continuity of care and limits opportunities for learning, follow-up and coordination between levels of the health system. The national referral and counter-referral guidelines explicitly recognize the need for feedback and appropriate documentation [5]. The deficiencies observed in the present study may therefore indicate an implementation gap rather than an absence of formal standards. Traoré’s study in Mali reported that counter-referral documentation was returned to referring facilities within one to two months in most cases [2]. Although the health-system context differs, this comparison illustrates that counter-referral mechanisms can be operationalized even in resource-constrained settings. Several factors may contribute to the weakness observed in the present study, including the absence of a clearly designated person or unit responsible for counter-referral, insufficient staffing, competing clinical priorities, inadequate documentation practices and the absence of an efficient mechanism for transmitting information to peripheral facilities. Strengthening counter-referral should therefore involve not only the provision of forms but also clear responsibility, standardized procedures, monitoring indicators and reliable communication channels [19].

Younger maternal age was associated with referral dysfunction. The original study interpretation suggested that younger women may be more dependent on partners or relatives when decisions involving referral and transportation costs must be made. Such an explanation is plausible, but the cross-sectional design does not allow this mechanism to be established directly. This finding should therefore be interpreted cautiously. Age may act as a marker for other social and economic factors, including parity, autonomy in healthcare decision-making, socioeconomic status or dependence on family members. Further studies incorporating these variables would be useful to clarify this association. Referral from a private facility was also associated with a higher prevalence of referral dysfunction. This finding may reflect differences in integration between private facilities and the formal referral network. Private facilities may have variable access to training, communication systems, referral protocols and institutional support. However, the present study did not directly measure these organizational characteristics, and therefore this explanation remains a hypothesis. The finding supports the need to ensure that referral standards apply consistently across both public and private health facilities. Training, supervision and integration of private providers into district referral networks may be particularly important.

Distance and rural origin were independently associated with counter-referral dysfunction, reinforcing the importance of geographical accessibility throughout the continuum of care. Maternal death was also strongly associated with counter-referral dysfunction. This association should be interpreted carefully because causality cannot be inferred from this cross-sectional analysis. In particular, counter-referral cannot logically be considered the cause of a death that occurred during the index hospitalization. Rather, the association may indicate that women who died had more severe complications, more complex care pathways or circumstances in which standard counter-referral procedures were not completed. This distinction is important because the observed association may reflect reverse causation or confounding by severity. Future studies should therefore examine whether characteristics of the referral process are associated with maternal death after adjustment for the severity and type of obstetric complication.

Taken together, the findings suggest that improving obstetric referral in northern Benin requires a system-level approach rather than isolated interventions. First, communication between referring and receiving facilities should be strengthened through reliable and low-cost communication mechanisms. Second, pre-referral stabilization should be systematically implemented, with clear protocols for emergency treatment before transfer. Third, access to appropriate transportation should be improved, particularly for women living far from referral hospitals [19]. Fourth, receiving facilities should develop mechanisms for rapid identification and reception of referred women. Finally, counter-referral should be incorporated into routine quality monitoring. Indicators such as the proportion of patients for whom feedback is transmitted, the completeness of counter-referral documentation and the proportion receiving documented follow-up recommendations could be regularly monitored. These interventions are consistent with the national strategy for reducing maternal and neonatal mortality in Benin and with the national referral and counter-referral guidelines [3][5][20].

In this study, the prospective data collection and exhaustive inclusion of eligible referred women during the study period reduced the risk of selection bias within the study population. The use of predefined referral and counter-referral criteria based on national standards also allowed the different components of the referral pathway to be assessed systematically. Several limitations should nevertheless be considered. First, this was a single-center, hospital-based cross-sectional study, which limits the generalizability of the findings to other health zones or referral hospitals in Benin. Second, some variables relied on information obtained from patients, relatives or referral documents and may therefore be subject to recall or documentation bias. Third, the use of a 70% compliance threshold was operationally useful but does not necessarily reflect the clinical importance of each individual criterion; failure to meet one criterion may have very different consequences from failure to meet another. Fourth, the observational design prevents causal inference regarding the associations identified. In particular, the association between counter-referral dysfunction and maternal death should not be interpreted as evidence that poor counter-referral caused maternal death. Finally, some variables had missing information, although these were retained as missing rather than replaced by assumed values.