Characteristics of included studies
Of the 17 articles included in the review, 11 utilised descriptive quantitative methods [1, 8, 33-41], four employed qualitative methods [42-45] whilst two used a mixed-methods approach [46, 47]. These studies are from an urban setting in Angola , rural setting in Burundi , South Sudan  and Ethiopia [41, 46]. Other studies were conducted in urban settings in Tanzania [36, 44, 45] and predominantly urban locations in Nigeria [34, 35, 48] and Ghana [1, 33, 37, 38, 42, 47].
Among the 15 facility based studies, six included only CEmONC facilities comprising the district, regional and teaching hospitals [8, 34, 37, 40, 44, 45]. Three of these studies investigated the quality of referral from BEmONC facilities, comprising primary and secondary level hospitals and health centres [33, 41, 46]. Five studies investigated health facilities providing CEmONC, BEmONC and normal birth service [1, 36, 38, 42, 47]. One study from Angola included peripheral birth units that could only attend to normal births . Health centres were described as facilities that provide normal birth service or provide less than six signal functions.
Most of these studies (n=13) investigated the quality of referral from the perspective of healthcare providers either through interviews or audit of health facility records [1, 8, 34-42, 45, 46]. Two studies explored women’s experiences of maternal and newborn referral [33, 44]. One study included both healthcare providers and women’s perspectives  whilst one South Sudan based study included the perspectives of stakeholders from the local government sector, Faith-Based Organisations (FBOs), non-governmental organisations (NGOs), and community religious leaders in addition to healthcare providers . The common conditions that prompted referral included premature rupture of membrane, obstructed labour and postpartum complications, such as haemorrhage and fistula [8, 37, 44-46]. The outcomes identified in these studies are summarised in Table 2.
Provision of maternity referral service
All the 17 articles reported on various aspects of maternal and newborn referral service provision and are presented under the following sub-themes as outlined by the supply side of framework for assessing the quality of maternal referrals (see Figure 1).
The characteristics of maternal and newborn referral systems were discussed in 12 papers [1, 33, 34, 36-43, 47]. Two studies indicated the use of standard maternal and newborn referral procedures, the availability of standardised referral forms and health provider escorts [1, 38]. Despite a lack of detail of these mechanisms, Awoonor-Williams, Bailey  indicated that such procedures were accompanied by telephoning ahead to prompt the receiving hospital about the referral.
In Ghana, one study that explored maternal and newborn referral among health workers in 16 facilities revealed that healthcare providers escorted all referred women and their newborns from health centres to the BEmOC and CEmOC facilities. However, details about the specific providers who escorted women and how this was done was not reported . Meanwhile, two studies investigating referral from CEmOC, BEmOC and facilities offering normal birth service reported that health care providers rarely escorted women to the next facility in Ghana [42, 47]. Some nurse administrators and medical doctors indicated that some referrals were unaccompanied because they were not emergencies . Women who were unaccompanied sometimes arrived late to the receiving facility, but details were not provided about the socio-demographics of these women or their circumstances .
Four referral pathways/routes were identified in the study by Elmusharaf, Byrne . These were zigzag referral (where a woman was sent back and forth between two healthcare providers), late referral to appropriate health facility, multiple referral (referring a woman to a number of non-functioning facilities before she finally arrived at the appropriate health facility) and by-passing (where health care provider at the first facility refers a woman directly to a specific higher level service in order to bypass non-functional health facilities). The by-passing pathway was found to result in timely arrival at the next facility. One woman narrated her zigzag referral:
Family members lifted the pregnant woman onto a donkey-driven cart and went to the village’s medical assistant. When they arrived, her water broke. The medical assistant prescribed drugs and told them that she was in labour and that the midwife should deliver her straight away. He sent them back to the midwife for birth. After spending three hours with the midwife without progress, the pregnant woman was exhausted. The midwife advised them to go back to the medical assistant. They spent most of the night going back and forth between a midwife and a medical assistant until the midwife insisted on the medical assistant referring them to Renk hospital (5JMD) .
Delay was described as a major feature of the maternal and newborn referral system. In one study from Angola delays of up to 13.7 hours were noted in the triage of a referred woman upon arrival at the receiving facility . Referred women and their newborns were simply added to the queue with other patients. When healthcare providers were alerted about this issue, a meeting was held to better identify referred women and prioritise them upon arrival. After this strategy was implemented, referred women and their newborns spent an average of nine minutes to meet a midwife and 71 minutes to be assessed by a doctor.
In Ghana, triage initiation for women in labour was found to vary by shift or timing of the day at Ridge Regional Hospital according to the study by Goodman, Srofenyoh . The average waiting time for triage in the morning was 35 minutes, 28 minutes in the evening and 55 minutes in the night, meanwhile, referral arrivals were even across shifts. The reasons for the disparity in the triage period were not indicated. Women in the first stage of labour were triaged within 35 minutes and moved to labour ward in 24 minutes afterward. Those in the second stage were evaluated in 30 minutes and moved to the labour ward within 10 minutes after triage .
One rural based study from Burundi found a 15% higher chance of neonatal deaths for women who took three or more hours to arrive at the next facility after they have been referred. The facilities did not have ambulances and had to call for an ambulance from another facility which sometimes delays . A study from Nigeria reported that 11 of 123 women died after being referred. Ten of the 11 women died of severe pre-eclampsia. Five of the women died during the postpartum period, five died during the intrapartum period whilst one died before labour commenced. Of these women, seven were referred from secondary health facilities, three from private hospitals and one from maternity home to the teaching hospital where the study was conducted .
Papers in the review identified various forms of transport that were used by women and their newborns to reach the facility to which they were referred. In the absence of ambulances, donkey-driven carts, tractors, pick-up trucks, public transport (taxis and lorries), motorised tricycles and motorbikes were used for referral purposes [33, 34, 38, 41, 42, 47]. Among the 720 confirmed pregnant women surveyed from five sub-districts in one study in Ghana, 15% took approximately one hour to make an arrangement to hire a private vehicle to transport them due to the non-availability of transport at health facilities . Ambulance use for all referrals from BEmONC or CEmONC services were implemented in a study from Ghana, but the resultant outcome was not reported .
Some health facilities in the included papers were not equipped with telephones and some healthcare providers from primary health centres in southern Ethiopia were described as having to use their personal phones to notify receiving facilities that women were being referred to them . Carnahan, Geller  described a communication system for facilitating referrals in 14 government health facilities in Tanzania. Meanwhile, details about how the communication systems enhance referrals were not explained. One study from a teaching hospital in Nigeria, however, reported the absence of initial communication to the hospital in all referrals . The survey did not report on reasons accounting for this practice.
Two studies indicated that referring health facilities are not always given feedback from the receiving facility so that opportunities for improvement can be identified [42, 47]. The high workload of health care workers was cited as the main reason why health providers did not receive any feedback about the outcome of women and newborns. A Medical Officer from Ghana reported:
‘‘We don’t receive any feedback from the hospitals. At least it would help us understand what we could have done better.’’ (In-depth Interview, Medical Ofﬁcer) .
However, one study indicated that feedback was sent to referring facilities and that this was provided verbally by the women and/or their families .
International standards for the management of emergency obstetrics
Five studies reported on the availability and or application of guidelines for maternal and newborn referrals [1, 38, 42, 45, 47]. Nwameme, Phillips  indicated that more than 75% of staff from two of the three facilities surveyed were trained on how to use the national referral guidelines. Shimoda, Leshabari  reported that midwives interviewed at one hospital and one health centre followed the guidelines of midwifery care in undertaking examinations leading to referral . However, the study did not identify the guidelines. Two of the three studies reporting the availability of the national referral guidelines in Ghana indicated adherence at health facilities [1, 47] whilst one noted non-adherence to the national referral guidelines . Adherence implies that healthcare providers initiate and execute referrals according to the steps outlined by the national guidelines. These include documentation of referral indication, telephoning the referral facility ahead of time and making arrangements for means of transport preferably ambulance . Kyei-Onanjiri, Carolan-Olah  reported that in Ghana, most of the 120 health facilities surveyed were guided by the national referral guidelines. For instance, 94% had standard referral forms and 83% followed a standard referral procedure .
In Tanzania, Shimoda, Leshabari  reported that guidelines for midwifery care were used by midwives' at the studied hospital and health centre in managing intrapartum monitoring processes leading to emergency referrals of 11 obstetric and prolonged labour. As compared to the WHO guidelines and recommendations [11, 49], the national guidelines of Ghana do not specify the uterotonics or oxytocics to be applied.
Human resource for health
Twelve studies reported on elements relating to the competencies of health providers involved in referring women and newborns to other facilities [1, 33, 35, 36, 38, 41-43, 45-48]. While a variety of providers were described as involved in maternal and newborn referral, few details were available regarding their experience and training. Three papers reported that midwives and nurses had between two and 12.5 years of experience in their positions [41, 45] and doctors and nurses with midwifery training had at least 6 years .
In the study by Carnahan, Geller , 17% of 115 healthcare providers (comprising nurses with nurses with midwifery training, nurses without midwifery training and medical/clinical doctors) surveyed from 14 government health facilities offering maternity services in Tanzania could correctly diagnose post-partum haemorrhage (blood loss ≥500 mL, or blood loss b500 mL with shock symptoms) . These facilities were urban CEmOC and BEmOC hospitals, health centres and dispensaries in Tanzania. Almost all the healthcare providers (98.3%) knew that misoprostol can be used for post-partum haemorrhage (PPH) but 62.6% of providers were able to state the recommended dose (600 μg) and only 36.5% were able to prescribe it because it was not always available . No significant differences were found in the mean score of the PPH-related knowledge index between providers who had more than six years of experience and those with less than six years of experience.
Women who had been referred to a higher level facility in one study from Ghana (76.6%) reported that healthcare providers at the receiving facility were competent enough to solve the problems for which they were referred, however, details about the providers’ competencies were not explained . In Nigeria, 79.0% (n=128) of Trained Birth Attendants (TBAs) surveyed in Kwara State indicated that they did not refer in a timely and appropriate manner . These TBAs were not formally trained but acquired their skills through either inheritance or self-initiation. However, TBAs (9 of 20) with at least a supervisory visit by qualified personnel were able to conduct timely and appropriate referral. Timely and appropriate referral involved the ability to refer high-risk pregnancies including women who had a previous stillbirth and women who were experiencing bleeding pregnancies. In addition, nine (N=13) TBAs who had attended more than one training course referred women with complications such as seizures, prolonged labour or retained placenta in a timely and appropriate manner .
Okafor, Arinze‑Onyia  also reported that some trained TBAs 155 (75.6%) in Nigeria, especially those who obtained the skill by inheritance delayed referral for women who showed signs of difficulty in childbirth by more than 12 hours . A senior manager at the Reproductive Health and Midwifery Department in the Renk County of South Sudan complained about the competency of TBAs:
In the past, TBAs have arrived in Renk hospital with pregnant women with their babies partly delivered; parts of the foetus, such as the head, the arm or the leg, outside the woman's body and the rest of the body still inside. (Senior Manager at the Reproductive Health and Midwifery Department) .
The poor skill set of lower level doctors and midwives were reported in Ghana:
“Last time a pregnant woman came here…. And I was saying but there is a doctor at your place, so why did you rush here without a midwife accompanying you, and she said ‘Auntie, I had been admitted there for a long time. And each time the doctor came, he said let's wait a bit more, and I was experiencing a lot of discomforts, and I insisted that they discharge me, so they ﬁnally reluctantly discharged me.' And when she arrived here, true, it was twins. But one was IUFD (macerated) already. So she was able to get the ﬁrst twin, but the second twin was macerated. (Midwife, District Hospital).
Health professionals noted the need for ongoing professional development. One midwife said:
“They (staff at the district hospital) need refresher courses... They should allow them to go to workshops so that they will see what is going on…. Me, I always learn from my junior nurses and midwives because I joined it [midwifery] about 10 years ago, and things are changing. Even the instrument[s] we are using [are] changing.” (Midwife, Health Center) .
Mirkuzie, Sisay  investigating obstetric referral in Addis Ababa, Ethiopia noted that midwives at an urban health centre, which is a BEmOC facility, promptly referred women whenever they identified obstetric complication induced by premature rupture of membranes (PROM):
“if a mother said that her water is broken, we consider her as PROM and we will immediately refer her to hospital for management” (an informant from HC A)
Maternity information systems
Five studies reported on maternity information systems [1, 37, 38, 42, 47]. Some studies from Ghana reported that some health facilities routinely used logbooks, care plans, referral letters, referral forms or slips correctly [1, 37, 38, 47]. Documentation was reported to be correctly done if all the necessary information about a woman were captured in the facility’s records as required by the national referral guidelines .
Meanwhile, poor documentation relating to referral was reported by one study from Ghana where six out of the 11 sampled health facilities had referral registers, which were documents for recording referral cases . Some referral cases and indications for referrals were improperly/incompletely documented. This documentation lacked details concerning the treatment offered by the referring facility, the current status of the referred woman and the required treatment at the receiving end [42, 47]. This resulted in instances where the accompanying staff member was unable to respond to specific questions about the care given before referring the woman. In the study by Nwameme, Phillips , only one EmOC facility among the three surveyed health facilities had a computerised referral information system.
Medicines and equipment
Of the seventeen included articles, four included information concerning medicines and equipment [1, 36, 39, 45]. Carnahan, Geller  reporting from Tanzania revealed that misoprostol was used for treating post-postpartum haemorrhage for received cases. However, this was not available in certain instances for some of these facilities that received referred cases . These facilities were urban-based CEmOC and BEmOC hospitals, health centres and dispensaries. A decline in correct use of partograph was reported from Ghana among some sending and receiving facilities by Awoonor-Williams, Bailey . The impact of this on referral was undisclosed by the study. An audit of periphery health facilities that provide normal birth service in Angola revealed poor quality of partographs . However, the reasons contributing to the poor quality as well as how these affected referral rates were not disclosed. From Tanzania, Shimoda, Leshabari  reported that catheters were correctly used to enhance referral decision making by midwives in monitoring intrapartum processes:
“When she put in the catheter, we saw some blood starting to pass. That is the sign of obstructed labour. That’s why I decided to refer immediately.” (F) .
Experience of referral service
Two of the included articles reported on two aspects of the experience of referral service [44, 47]. These aspects are human resource for health and satisfaction with facility, commodities and cost. None of the articles included in this review provided insights into cognition and emotional support aspects of referral that formed part of the conceptual framework.
Human resource for health
In a study from Tanzania, some women in a primary health facility indicated that doctors were not readily available to check their progress and then refer them to the facility where EmOC can be provided if necessary . One woman narrated her story:
“When we got to the dispensary nurses told me to wait. At 8 pm labour pains became intense, I started pushing but the baby could not come out, and the doctor was not around. Next day I continued pushing the whole day again until at around 8 pm when the doctor came…” (Divorced, aged 33, MboriDodoma), .
The poor attitude of nurses was reported by some women. In Accra Ghana, just over 10 percent (N= 390) of women at three health facilities (two health centres and one polyclinic) indicated that poor attitude of nurses was a source constraint to referral . However, the specific attitudes were not disclosed by the study.
Satisfaction with facility, commodities and cost
Only one study reported on the satisfaction women had for the services they received or should have received . Sixteen of 390 women surveyed at three urban facilities that provide BEmOC and normal birth services were dissatisfied with the performance of the health facilities. Another 180 of 390 complained about the cost whilst three reported previous bad experiences as sources of dissatisfaction and constraint to referrals. Nwameme, Phillips  did not provide details regarding the specific aspects of the facility that affected their satisfaction.
Socio-cultural factors affecting women’s adherence to referral
While the framework did not include socio-cultural factors, these were identified as having an effect on a woman's referral to another facility for care. In the study by Afari, Hirschhorn , health care providers in Ghana reported that socio-cultural beliefs relating to fear of blood transfusion and fear of death at higher level facilities affected women’s desire to travel to the next level facility. Nuamah, Agyei-Baffour  report that 57% (N=720) of the women they surveyed at a CEmOC health facility in Ghana had to speak to their husbands before they were able to follow the advice of health professionals . According to Mselle and Kohi , some women in rural Tanzania who were suffering from obstetric fistula indicated that the decision to travel to the next facility was made by their uncles, grandmothers, husbands and mothers-in-law.