Assessing the Functionality of an Emergency Obstetric Referral System Among Public Healthcare Facilities in A Low Resource Setting: An Application of Process Mapping Approach

Background: Weak referral systems remain a major concern inuencing timely access to the appropriate level of care during obstetric emergencies, particularly for Low-and Middle-Income Countries, including Ghana. It is a serious factor threatening the achievement of the maternal health Sustainable Development Goal. The objective of this study is to establish process details of emergency obstetric referral systems across different levels of public healthcare facilities to deepen understanding of systemic barriers and preliminary solutions in an urban district, using Ablekuma in Accra, Ghana as a case study. Methods: The study is an analytical cross-sectional study. Nine (9) in-depth interviews (IDI) were carried out for a three-weeks period in June and July 2019 after informed consent with two (2) Obstetrics & Gynaecology consultants, two (2) Residents, one family physician, and four (4) Midwives managing emergency obstetric referral across different levels of facilities. Purposeful sampling technique was used to collect data that included a narration of the referral process, and challenges experienced with each step. Qualitative data was transcribed, coded by topics and thematically analysed. Transcribed narratives were used to draft a process map and analyze the defects within the emergency obstetric referral system. Results: Out of the 33 main activities in the referral process within the facilities, the study identied that 24 (73%) had a range of barriers in relation to communication, transport system, resources (space, equipment and physical structures), stang (numbers and attitude), Healthcare providers (HCP) knowledge and compliance to referral policy and guideline, and nancing for referral. These ndings have implication on delay in accessing care. HCP suggested that strengthening communication and coordination, reviewing referral policy, training of all stakeholders and provision of essential resources would be benecial. Conclusion: Our ndings clearly establish that the emergency obstetric referral system between a typical teaching hospital in an urban district of Accra-Ghana and periphery referral facilities, is functioning far below optimum levels. This suggests that the formulation and implementation of policies should be focused around structural and process improvement interventions, strengthening collaborations, communication and transport along the referral pathway is likely to ensure that women receive timely and quality care. mapping developed from the perspective of frontline health workers to examine the functions of an emergency obstetric referral system in a Low-and Middle-Income country. It breaks down the complex steps of the overall referral system as a woman presenting with emergency obstetric complication transits from a lower level to a higher-level public healthcare facility highlighting the barriers and preliminary solutions that can support future policy and improvement initiatives.

different levels of public healthcare facilities to deepen understanding of systemic barriers and preliminary solutions in an urban district, using Ablekuma in Accra, Ghana as a case study.

Study design
This study employed a cross-sectional design approach. This study adopted the Interpretive/Constructive paradigm to understand how the referral system operates as opposed to how it is intended. The researchers of this study recognised that frontline health workers and clinical leaders from varied backgrounds have their perspectives, beliefs, assumptions, and experiences that would contribute to the reality existing about the broader functionality of the referral system.

Study setting
This study was conducted in three public health facilities in the Ablekuma district, speci cally: in the Obstetrics and Gynaecology department of Korle Bu Teaching Hospital (KBTH) the third largest hospital in Africa and the largest tertiary hospital in Accra, Ghana; Mamprobi Polyclinic (MPC); and the Dansoman polyclinic (DPC). The setting was purposively selected to include public health facilities in an urban setting that providing primary, secondary and tertiary healthcare in the capital city of Ghana, Accra.
With regards to healthcare coverage, the Obstetric unit of KBTH provides specialized antenatal care, postnatal care to large and diverse population within and outside Accra and from lower levels of facilities primary healthcare facilities and district (secondary) hospitals both public and private. The total bed capacity for the hospital is over 2500 with 375 bed capacity for maternity services (27). The hospital runs an antenatal clinic, has two labour wards and three theatres.
The KBTH handles an estimated 27,128 new and old antenatal (ANC) attendance, 16,000 postnatal attendance and over 10,000 deliveries every year. About 70% of maternal deaths are referred cases (26).
For MPC and DPC, they re ect a cross-section of two levels of healthcare, primary and secondary. The KBTH and MPC are located in Ablekuma South whilst DPC is located in Ablekuma West. The map of the Ablekuma district is shown in Figure 1. The DPC runs only antenatal clinics with one couch for ANC, 8 adult beds and 2 cots, a total capacity of 10 beds for the entire polyclinic. The Mamprobi Polyclinic runs an antenatal clinic, has a labour ward and an operating theatre. The annual ANC attendance in MPC is 18,677, the annual PNC attendance is 3089, annual deliveries is 2,678 annually and a total bed capacity of 45.
The Ablekuma South district has an estimated population of 213,914 (25). The KBTH is about 2.66 kilometres from MPC and about 6 Kilometers to DPC. The distance between DPC and MPC is about 4.65 kilometers.

Data collection
Healthcare providers executing varied professional roles were recruited using purposive sampling technique to obtain in-depth knowledge, individualized experience, and perception of the emergency obstetrics referral system. Data was collected using semi-structured key informant interview guide (see IDI Referral Interview Guide) designed for this study to gather feedback about the process ow, existing barriers and recommendations for improvement. Data collection was undertaken by the primary researcher (BO) who has experience in undertaking qualitative public health and clinical research conducted nine indepth interviews (IDI) for a three-week period between June and July 2019 until saturation was met.
Participants were approached face-to-face by the data collector (BO) prior to the interview and were given written and verbal information about the study. After securing their written consent to participate in the study, a date, place and time was scheduled for the interview. All HCP approached accepted to participate in the study.
Two Obstetrician and Gynaecology consultants, two OBGY resident doctors, one Family Physician Specialist, and four Midwives across the three facilities in Ablekuma district in Accra, Ghana, were interviewed in English. Five were staff from KBTH, two from MPC and two from DPC. Interviewing involved the act of asking the respondents questions and audio-tape recording the responses and transcribing upon completion. In-depth interviews lasted on average fteen minutes. All interviews were conducted in person in a private o ce in the healthcare facility that would ensure privacy and convenience for the participants.
The transcripts were randomly checked against the audio recordings for quality assurance purposes.

Data analysis
A qualitative analysis was conducted in ve phases. First, one researcher [BO] manually transcribed the interviews. This was reviewed by two other researchers [DO, MN] to eliminate bias, ensure consistency and check reliability. Second, we independently identi ed referral process related data in breadth and depth, from all nine interviews. Third, we used the transcribed narratives to draft a process map for the current emergency obstetric referral system within Ablekuma district, for each of the interviews. This was done initially for the lower-level facilities (MPC and DPC), then the higher-level facility (KBTH). Fourth, we re ned and con rmed the process map ensuring that details from all interviews were re ected. Particular attention was paid to uncertainties about the referral process between the speci ed start and end points. That is, from when a woman with an obstetric complication comes to a lower level facility to when discharged from a higher-level facility respectively. Finally, the developed process map provided explicit visualisation of the referral process which was further analysed in two steps. First, we examined the uncertainties in the sequence of steps within the process and/or lack of systematised steps, gaps [that is, discrepancies of what the process is intended to be and what is actually is], bottlenecks within the process that cause delays before the next step occurs and ine ciencies [unnecessarily repeated steps leading to delays] to uncover potential areas for improvement broadly similar to approach used by (28). Key themes and sub themes were also discussed and reviewed by the ve researches [BO, DO, MN, KAP, TB]. A meeting was later held with the participants to con rm the ndings.

Results
A total of nine (9) healthcare providers participated in the study: two Obstetrician and Gynaecology (OBGY) consultants, two OBGY resident doctors, one Family Physician Specialist, and four Midwives across the three facilities in Ablekuma district in Accra, Ghana. Five of the staff were from KBTH, two from MPC and two from DPC. All interviews were carried out in English. Overall, the main themes derived from the data analyses were (1) process ow/map (2) barriers (3) healthcare provider recommendations. The variation in the cadre of healthcare provider did not impact the response patterns.
Process Flow/Map Figure 2 above presents a process map designed on the basis of frontline healthcare workers' narratives. It articulates the steps involved during a referral for women with emergency obstetric complications from lower level facilities to a higher-level facility in Ablekuma district.
Lower level facility: The referral process in this study starts when a lower level facility receives a patient [a], they then make a diagnosis [b], stabilises the patient (c) and make a decision to refer [d] or not [e]. If a decision is made to refer, the referring facility prepares the patient and relatives [f,g], calls or noti es the receiving institution [h] and the ambulance service providers. The ambulance service again calls the receiving institution to con rm their readiness to accept the case and also con rms with the referring facility the patient's ability to pay with the referring institution [i]. When the higher level facility gives the green light, the lower level facility completes lling the referral form and document in the referral register [j], The ambulance service driver then moves the ambulance from its station to the referring facility (MPC or DPC), then to the receiving facility accompanied either by patient relatives, a midwife, a nurse, a healthcare assistant or student nurse (s) [n] whilst the patient is being continuously monitored [o, p]. In a situation where the ambulance service is unavailable, a taxi or a private car of relatives is used to carry the patient to a higher-level institution (KBTH). A decision is made on whether a relative or a healthcare worker should accompany the patient [l].
Higher-level facility: The higher-level period involves the following activities:

Barriers in the referral System
Out of the 33 major points/steps, ndings suggest that 24 steps [c, d e, h, i, j, l, m, n, o, p, q, r, s, t, u, v, w, y, z, aa, dd, ee, gg; noted using coloured coded shapes] often led to delays in moving from one step to the other while transferring the patient to an appropriate level of care, shown in Figure 2. Barriers observed stepby-step in the referral process is detailed below:

Stabilizing the client [c], deciding to refer [d], and manage internally [e]
In stabilizing the patient at the referring facility, healthcare workers at Dansoman Polyclinic (DPC) mentioned that there was no labour ward or a theatre at their facility. They had very limited space and lacked certain essential emergency medicines to provide care for women with emergency complication (s). Unlike DPC, Mamprobi Polyclinic (MPC) had a theatre, a labour ward and a space for antenatal clinic. However, health workers at MPC reported that there was no dedicated Obstetrician Gynaecologist (OBGY) readily available to undertake emergency surgeries should there be a need for one, at the time of collecting data. The MPC thus, relies on external OBGYs mainly from Korle Bu (the higher-level facility) to perform the surgeries in Mamprobi; that is if these OBGYs are available. Otherwise, MPC refers to a higher-level facility for further management. This situation often led to delays in deciding to refer the patient.

Alerting receiving facility [h]
It was evident that staff from lower-level facilities feared their patients would be rejected when they call in advance to notify higher-level facilities. The fear and perceived bureaucracy deterred healthcare workers [e.g. midwives] at lower level facilities from notifying the receiving facility in advance before sending patients. One doctor from the receiving facility noted: "…, some of these facilities just send patients without informing the recipient facility because sometimes the recipient facility do not have space, do not have bed and they think that if they are told there is no bed, they would have to contend with the di culties that they are not prepared for, so they would rather want to stampede and just come without informing the recipient facility…" A2 Some healthcare workers from one of the lower facilities also noted: "…sometimes you call and call [the receiving facility] and no one picks up the call …" A9 "…Most at times too, the call doesn't go through and we are forced to move the client without informing a particular receiving end and you will be turned back because you haven't called and this client may lose her life …sometimes if you don't meet the one you spoke to on the phone the case may be rejected with the excuse that they were not informed. This usually happens during shift hours…" A8

Arranging for transport [i]
The process involved in arranging for a means of transport for referrals for women with emergency obstetric complications seemed cumbersome. Additionally, there seemed to be a lack of trust among the referring institutions and stakeholders. For instance, the ambulance service providers always con rmed that the referring facility had called the higher facility even after they have been informed by the referring facility that they had already called. This led to bureaucracy and delay in dispatch of an ambulance to the facility that needs to refer the patient. As noted by a midwife: "…The ambulance is also an issue because they always delay the process because they also con rm whether the client will be able to pay for the service before they call you back to inform you that they have con rmed from the receiving end before you can prepare for the transfer…"A7 Another midwife described a scenario: "…So, you have called whichever hospital you want to refer to and after they have accepted then you call an ambulance. The ambulance will also ask if you have already called the receiving end after which they [ambulance service providers] will call the receiving end to con rm before they accept to come…" A8 Financing for ambulance services was also a challenge. Speci cally, the cost involved when using an ambulance was relatively high. Most patients who needed an ambulance did not have the ability to pay. One midwife described the situation. She said: "…sometimes getting an ambulance is a bit of challenge. You have to resort to the private ambulances and some of them do charge fees at higher rate …so if we assess the patient and she cannot afford, then we use the taxi." A9

Filling of referral form [j]
It was observed that a standard referral form was used by the referring facilities. However, respondents from a receiving facility noted that most often, lling of the referral form is incomplete which affected continuity of care. It was therefore di cult to seek clari cation on patient's prior history or management. One doctor at the receiving facility narrated: "…sometimes where the referral details are scanty, you may want to get in touch with the referring facility to get them to ll in the gaps. But it doesn't happen because often we don't have the numbers on the referral form…" A2 Deciding if patient condition is critical [l] to Accompanying patient [n] It appeared that the assessment of patients' condition before a decision is made to accompany patient or not was not optimum. Consequently, respondents from the receiving facility expressed their displeasure at seriously ill patients coming in unaccompanied or accompanied by a person who could not provide them with the needed information about the patient. This was con rmed by one of the referring facilities that, due to their numbers, they assess to see the seriousness to join the patient to the referring facility or not. The study gathered that unsuitable means of transport: taxi and private cars were frequently used to transport patients from a lower facility to a higher-level facility due to their availability. The healthcare workers reported that this affected adequate monitoring and proper handling of patients whilst in transit to a higher facility. Consequently, such patients arrived at the higher facility in a poor state of health and in some instances, some patients lose their lives before arrival. One midwife described the effect of the non-availability of the appropriate means of transport. She said: "…let's say you are coming with a taxi and a client needs oxygen. Automatically, there is no way you can give oxygen to a patient in a taxi especially if they refer without an ambulance…." A3

Arriving at receiving unit (KBTH) [q] and Triaging of patients [s]
Despite KBTH having a patient triaging system, almost all respondents in the receiving facility admitted that the structural design at their reception made it di cult for patients, their relatives and/or a new nurse to nd the emergency area. They explained that relatives or accompanying midwives had to pick up a wheelchair and transport patient to the emergency. In some instances, patients who do not know what to do walked to the emergency or triage area; which was about 50 meters from the entrance of the maternity unit. In a case where an ambulance is used, the patient is wheeled to the emergency. According to a respondent: "…the other bit is lack of well written out guidelines pasted at the emergency and entrance of the facility. Because of this gap [no notices at the entrance] people by their own way either carry the patient to the emergency or if they have a wheelchair at the entrance, they basically put the patient in it to the emergency…" A2 Similarly, it was gathered that the workload at the receiving facility did not allow healthcare workers to sometimes attend to cases speci ed in referral notes as emergencies as recommended by the referral policy, leading to delay in providing care to the patient. As one midwife narrates: "…a case I see as an emergency may not be seen as an emergency at the referral facility. The client may have to join the queue when she gets there and this may worsen their condition…" -A9

Handing over with referral form and patient documentation [r], and Debrie ng receiving midwife/doctor about patient [t]
Handing over of patient from the referring facility to the receiving facility was a challenge in some instances due to incomplete lling of referral forms, incomplete patient documentation and patient being accompanied by persons who do not have su cient information. However, it seemed prior notice to the higher facility facilitated smooth handing over. As narrated by one midwife: "…If they are aware that this patient is coming, they will receive but if they are not, they will toss you around and eventually you will have to return with the patient…" A6

Initial resuscitation [u]
It is assumed that KBTH as a tertiary facility should have adequate manpower, equipment and other resources. However, issues of shortage of equipment and supplies existed. In addition to this, staff attitude affected the higher-facility's preparedness and responsiveness to providing initial resuscitation. For instance, a midwife at the receiving facility mentioned that: "…the things [Equipment] are there but as to whether they are working by the time you will take up they don't check. For instance, oxygen cylinder it is there, the ow meter and everything but they won't check so a patient may come needing oxygen and they open and it is empty …" A4

Communication with senior person/consultant [v]
It appeared that the infrastructure or gadgets supporting communication in these facilities were not e cient and reliable. In some instances, healthcare workers had to use their own mobile phones to make calls. One doctor mentioned that: "…Gutta phones existed but they are not functioning now …, so the information ow is not as it should be for such patients…" A1

Additional investigations if any [w]
The National Health Insurance scheme (NHIS) was reported to be ine cient by three interview respondents. This according to them delayed urgent diagnostic investigations.

Sending patients to theatre [y] and Recovery [z]
At the receiving facility, two theatres were functioning. The third theatre was not functioning, due to the unavailability of patient monitors. As explained by a respondent at the receiving facility, an average of 11 caesarean sections and about two (2) emergency surgeries such as hysterectomies were done in a day.
Averagely, a caesarean section and cleaning up took about an hour and half while a complicated obstetric surgery and cleaning up took about three hours. Importantly, about 30 women were delivered in KBTH in a day. About forty percent of these women often required caesarean section. Meanwhile, there were only four beds in the recovery ward and in some instances one or more of these beds were occupied by a seriously ill patient for days.

Document in referral register [dd]
At the receiving facility there was a book that captures all daily admissions to the hospital. This captured all cases whether referred or attendant at the KBTH. This suggested that no register was dedicated exclusively for referred cases. This made it di cult to easily retrieve information concerning referred cases.

Give feedback to referral facility [ee]
Giving and receiving feedback allows for learning and continuous quality improvement. However, it seems to be non-existent as one doctor describes it: "…feedback is virtually none existent until now that a rapid response team even came up and that is done at random. It is with every case that is referred that there should be feedback. But that is not happening at all. So, the feedback is poor just as the referral itself …." A1 In the receiving facility, there seemed to be no dedicated people who had complete information about all referred patients. This made referring facility personnel who tried calling to follow-up frustrated. Unless informal calls were made to personal colleagues to get feedback in higher-level institutions or calling patients directly, the feedback was not obtained.
One midwife and a doctor from a referring facility respectively mentioned that: "…the higher facilities do not call to inform us about the cases we refer to them. No, it doesn't happen like that unless the patient has passed away then you will be called and informed of what had happened… it is the relatives or clients we usually call to get feedback from and not the facility…" A8 "…On our part yes, we seek feedback but let me say that it hasn't been formalized …." A7 Some of the healthcare workers from a referring facility attributed the non-existence of feedback to indiscriminate referrals which led to workload. It also appeared that healthcare leadership's efforts in ensuring that procedures and processes were followed were inadequate. This affected the attitude and motivation of staff in making efforts to give feedback as they perceived nothing will be done to improve the process. One doctor gave a reason for lack of feedback: "…Part of it may be that when you spend time to do proper feedback it may not actually come to anything because there isn't any channel of command on the ground to ensure that things are looked at the way it should be looked at and things are corrected the way it should be corrected. No audit of what is going on… What case do you refer? What is the response? There is nothing happening as I am aware of in the referral sources to push anyone to look at what you have written. There is even not a proper channel. the referral system is not working the way it should work. …the chain of command on the ground is not being adhered to. And people are not being held or sanctioned for not taking responsibility …" A1

Referring stable patients back to lower facilities [gg]
Referring stable patients from higher-level facility back to lower-level facility seem to be non-existent. As one doctor noted: "…when patients are referred and they come with certain conditions and they are stabilized, it should be possible for them to be referred back to their primary healthcare facilities…... But that does not also happen so usually you have all of them eventually coming to Korle Bu …"-A2

Other Barriers
Other barriers which affected the process were also identi ed. These were mainly related to the referral policy and guidelines; and leadership and Accountability.
i. Non-availability of policies and procedure documents: Referring institutions had displayed adopted referral guidelines and contact numbers of referral focal persons at vantage points. However, guidelines on obstetric referrals including emergencies were not visibly displayed in the receiving facility. This questioned if a well laid down process existed. These concerns were con rmed by statements made by two doctors in the receiving facility: "…Yes, there is a written book and policy on the referral processes, except that I will not be able to lay hands on that book now. It is there [referral policy] but unfortunately looking from then on it is not being followed because of the high turnover of residents and nurses so at the point they will not even remember there was something like that..." A1 "...we don't have a well laid out process a number of the things are by convention so most of the time, people are told that this is how the emergency functions when they come in, but we don't have a well laid out thing which is probably written and pasted at vantage points for people to know that this is how the emergency is supposed to run…"A2 ii. Insu cient knowledge of policies by health workers: It also appeared that healthcare workers did not have su cient knowledge about the existence of referral policies. One midwife mentioned: "…Well, I don't know if there is any protocol. All I know of is there is an exchange center where you call to nd out if there is a space. Apart from that, I don't know of any other…" A3 iii. Leadership and Accountability: "…But the chain of command on the ground is not being adhered to… So, it [Referral System] is not working the way it should work…" A1 Overall sub-themes that emerged during qualitative analysis as barriers encountered in the emergency obstetric referral system were in relation to: communication, transport and nancing for referral system, resources (space, equipment and physical structures), sta ng (numbers and attitude), referral policy and guideline.
Critical analysis of the process map revealed gaps, uncertain steps, ine cient steps, and bottlenecks. In terms of gaps, the referral process described by respondents broadly conforms to requirements in Ghana's Ministry of Health Referral Policy Guideline document (21). However, there were some discrepancies between what this guide is intended for and how the process is carried out. For instance, the two-way referral from higher facility to lower facility rarely . Similarly, some steps in the process were unnecessarily replicated. For instance, in situations where the midwife needs to refer a client, she calls the ambulance service to notify them about a patient needing transport. Likewise, she is required to call the higher-level facility to enquire about their readiness. Meanwhile, to con rm this, the ambulance driver also calls the higher-level facility to enquire about their readiness to receive the patient [h].

Suggested Recommendations
The HCP shared their views on what strategies could help improve the current emergency obstetric referral system. They suggested an audit of the current referral forms; organisation of referral focused meetings, assigning a trained and dedicated personnel to effectively coordinate and monitor communication (  "…there needs to be a better coordination between the referral sources and experts within the Ghana health service as well as outside so cases are not just pushed unnecessarily as an emergency to clog up clinics and clog up the wards that are required by genuine emergencies…" A1 Doctor Q8 "…getting the referral forms in such a way that referral facility eventually after the care has been given can also write back to the referring centres … because patients are stable…" A2 Doctor Q9 "…certain consumables… and the health insurance system should be reviewed and well-coordinated to help patient get the best care at the shortest time …A1Doctor Strengthening Transport System Q10 "…If ambulance service would overlook where they have to con rm from the receiving end before attending to the emergency will help to minimize the delay process. Q19 "…Redesigning the entrance …once a vehicle lands there should be someone that people can approach ….and do resuscitation along the line…." A2 Doctor Q20 "…the facility [DPC] must get its own delivery ward, get an attached obstetrician if not permanent one we can easily call to assist when we have problem and the emergency bed management improve at the higher facilities…" A6 Doctor Q21 "…it will go a long way to help if at least if the major once gets their own obstetricians... The person will not be able to work 24/7 and will lessen the number of referrals…" A7 Doctor Q22 "…most drugs for pre-referral treatment be made readily available if not free and also make available logistics to work with…" A9 Midwife

Discussion
In the midst of global efforts to achieve the SDG related to maternal health, identifying improvement opportunities has been of much interest (10,29). This study utilised process mapping developed from the perspective of frontline health workers to examine the functions of an emergency obstetric referral system in a Low-and Middle-Income country. It breaks down the complex steps of the overall referral system as a woman presenting with emergency obstetric complication transits from a lower level to a higher-level public healthcare facility highlighting the barriers and preliminary solutions that can support future policy and improvement initiatives.
The results of our study reveal that the referral processes for both lower level facilities were similar and the sequence of steps was generally logical. However, HCP thought that the current emergency obstetric referral system is not working as it should. Across the 32 major steps taken during emergency obstetric referral in the district, on average almost three-quarters of the steps had defects, and critical steps like stabilizing patients at the lower level facility and initial Ethics approval and consent to participate Ethical clearance was granted by the KBTH Institutional Review Board, Accra Ghana (Protocol ID NO: KBTH-IRB/00099/2019). Administrative approval was obtained from the Accra Regional Health Directorate to pretest in Ussher polyclinic and collect data in Dansoman and Mamprobi polyclinics. Verbal informed consent was obtained from all the respondents. Con dentiality, anonymity and privacy of respondents was maintained at all times. The researchers removed names and other identifying information in the responses to ensure anonymity, and this was explained to respondents before seeking their consent. Study participants were also given the right to withdraw and this could be exercised at any time during the study and debrie ng. Data les were password protected and hard copies locked and access was limited to the three main authors and the data analyst of the study. Data is being used for research purposes only. The researchers have no con ict of interest so far as this study is concerned.

Consent for publication
Not Applicable Availability of data and materials The datasets analyzed in this study are available from the corresponding author on reasonable request.