Introduction

Ghana has made significant progress towards promoting access to obstetric services as reflected by a rapid increase in the institutional delivery rate over the past decade. Currently, 79% of expectant mothers deliver in health facilities assisted by skilled birth attendants, an increase from 54% 10 years ago [1]. Despite this improvement, significant gaps remain in the provision of quality care for high-risk obstetric patients, particularly along the referral pathway [2–6]. At least 15% of pregnancies develop complications that require advanced care, including surgery, in hospitals capable of performing comprehensive emergency obstetric care (CEmOC) [7]. The World Health Organization currently recommends that emergency obstetric services should be available within two hours of seeking care [8]. Similarly, the Lancet Commission on Global Surgery recently adopted six quality of care indicators to assess surgical readiness, among which is access to care within two hours to facilities capable of performing emergency surgery [9].

Delays in reaching appropriate health facilities contribute to poor maternal and newborn outcomes [10–13]. Inefficient referral processes can result in death during transport or soon after arrival at secondary and tertiary institutions [2, 3, 13]. As such, referral and teaching hospitals have maternal mortality rates that far exceed the national average [3, 14, 15]. In Ghana, the country action plan for the Millennium Development Goal (MGD) Acceleration Framework identified weak referral systems, particularly for managing obstetric emergencies, as one of the leading challenges to achieving the MDG for maternal mortality [16].

Numerous measures to strengthen Ghana’s referral system have been undertaken. In 2012, the Ministry of Health deployed a national policy to address delays in accessing emergency care for referred patients [17]. Other measures have included a National Health Insurance Scheme with free access to maternity care, a national ambulance service, implementation of maternal death audits and the use of audit findings to inform referral institutions, the establishment of referral ledgers, and policies to ensure that health staff accompany referred emergencies [2–4, 17, 18]. In addition, the Greater Accra Region has established a “call centre” to facilitate linkages between the source and receiving hospitals. Despite these interventions, the referral system remains challenged [2]. There are recent reports in the lay press on the “no bed syndrome” whereby referred emergencies are denied admission and care at one institution and are repeatedly referred elsewhere [2].

A way to improve the referral process is to leverage messaging technology to promote collaboration between hospitals. Smartphones, tablets, and messaging services have become commonplace. WhatsApp is a popular messaging service with over 2 billion users in 180 countries [19]. It is a free smartphone application that utilizes the internet to share text, photographs, video, documents, and voice calls. With end-to-end encryption, WhatsApp is an emerging medium for healthcare providers, especially in low resource settings, where access to landlines, computers, and other communication tools is limited [20, 21].

A feasibility project was undertaken in Ghana to utilize WhatsApp as a mechanism through which the high-risk obstetric referral process could be improved. Conceptualized by local physicians frustrated by the ineffective referral process, the idea was to create an inter-institutional communication platform that would link healthcare providers at primary level health centers and district hospitals to senior clinicians at a large regional hospital in the Greater Accra Region. The goal was to improve communication among clinicians to help identify cases appropriate for referral, guide treatment interventions, expedite inter-institutional transfers, facilitate feedback, and improve system-level accountability by including hospital and health service administrators and the national ambulance service. The intervention was thought to be a viable, low-cost solution to referral communication difficulties since smartphones and WhatsApp are commonly used throughout Ghana. Reports on the use of WhatsApp in maternal and child health literature have been limited. WhatsApp has been used to facilitate pregnancy support groups in Kenya [22], to enhance communication among maternal-fetal medicine specialists across continents [20] and to support neonatal referrals in Cameroon [23]. However, to the authors’ knowledge, only one small study exists that evaluates the use of a mobile messaging application on addressing CEmOC referral care [24]. The aim of this study was to determine the feasibility of using WhatsApp as a communication tool among clinicians as well as to understand timeliness and other characteristics of obstetric referrals to a large regional hospital in Accra, Ghana.

Methods

Results

Characterizing referrals to GARH

The GARH had 1266 referrals from September 1 to November 30, 2017. Data were collected on 652 of these; however, 55 were excluded due to incomplete referral and arrival times. Data were analyzed for 597, representing 47.5% of referrals during the study timeframe; 319 (53.4%) were from Platform and 278 (46.6%) were from non-Platform institutions. There were 370 (62.0%) referrals from health centers; 85 (14.4%) from district hospitals; 16 (2.7%) from tertiary or teaching hospitals; 120 (20.3%) from private hospitals, and 6 (1.0%) were from unknown facilities. Six (1.0%) patients had two referral points. Referrals emanated from 114 institutions; Fig 1 shows the locations of the ten highest volume source hospitals, 8 of which were Platform institutions. More referrals were made and received during the day (8:00–19:59); however, a significant proportion of these arrived at night (20:00–7:59) [referral time: day 444 (74.4%), night 153 (25.6%) vs arrival time: day 371 (62.1%), night 226 (37.9%); p<0.001)]. This was consistent across Platform and non-Platform facilities.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Map of Accra, Ghana, demonstrating the location and volume of obstetric referrals from the ten leading referral facilities to the Greater Accra Regional Hospital.

The figure was created using Tableau data software and the map was generated using Mapbox open-source utilities. The Map image used open-source software found at https://www.mapbox.com/about/maps/ [mapbox.com], with data from http://www.openstreetmap.org/about/ [openstreetmap.org]. Everyone is welcome to improve the map here: https://www.mapbox.com/contribute/[mapbox.com].

https://doi.org/10.1371/journal.pone.0266932.g001

Patient demographic information is shown in Table 1. The distance traveled, mode of transportation, and accompanying personnel are shown in Table 2. Platform hospitals were significantly closer in proximity to GARH than were non-Platform hospitals (p<0.01). The most distant referral site was 145.8 km from GARH. As shown, taxi was the predominant mode of transportation. There were no differences in mode of transport among facility types.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Patient data and education of referred obstetric patients.

https://doi.org/10.1371/journal.pone.0266932.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Distance, mode of transportation and accompanying person.

https://doi.org/10.1371/journal.pone.0266932.t002

Labor characteristics and referral indications.

There were 261 (43.7%) antenatal, 336 (55.9%) intrapartum, and 2 (0.3%) postpartum referrals, the distribution of which was similar across Platform and non-Platform sites. The indications for referral are shown in Table 3. The majority of referred patients arrived with a referral slip (404; 67.7%); this did not differ between Platform (227; 71.2%) and non-Platform (177; 63.7%) facilities. In the majority, a pre-referral phone call was not made [overall 204 (37.2%); Platform 115 (39.2%); non-Platform 89 (34.8%)]. Pre-referral treatment was significantly more common among Platform facilities, albeit low [overall 129 (25.1%); Platform 81 (28.5%); non-Platform 48 (21.0%) (p = 0.05)]. Among the referred patients, 588 (99.8%) delivered at GARH; 371 (63.4%) had vaginal and 214 (36.6%) had cesarean deliveries, consistent among Platform and non-Platform referrals.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Indications for referral.

https://doi.org/10.1371/journal.pone.0266932.t003

Urgent referrals.

Five hundred fifteen (86.3%) patients were deemed urgent referrals for emergency conditions; the most frequent were labor dystocia (189; 29.8%), hypertensive disorders of pregnancy (90; 14.2%), prior uterine scar (59, 9.3%), maternal comorbidities (46, 7.3%), and fetal compromise (42, 6.5%) without variation by facility type. Urgent indications accounted for all but one of the midwife accompanied referrals shown in Table 2. The mode of delivery for urgent cases did not differ from that stated overall.

The median (IQR) referral to arrival time for urgent referrals is shown in Table 4. When data were disaggregated by mode of transportation, arrival time was significantly shorter for those arriving by ambulance (p = 0.018). Referrals that were accompanied by a midwife also resulted in significantly shorter median referral to arrival times in Platform (60 min) vs non-Platform (135 min) hospitals (p = 0.003). Only 140 (23.5%) patients referred urgently arrived within 2 hours with 85 (30.1%) from Platform and 55 (22.7%) from non-Platform sites (p = 0.059). Of the 74 urgent referrals accompanied by a midwife or nurse, 44 (59.5%) arrived within 2 hours.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Referral to arrival time for urgent referrals.

https://doi.org/10.1371/journal.pone.0266932.t004

Outcomes observed in referral analysis.

Maternal and fetal outcomes are shown in Table 5. There were 30 stillbirths; four were known prior to transfer, 10 occurred during the referral process, 14 had a fetal heartbeat on arrival at GARH and in two the fetal heart rate was not recorded. In these, the referral to arrival times varied widely, from 20 min to several days. One woman was referred for failure to progress and fetal distress; however, on arrival she was diagnosed with uterine rupture and intrauterine fetal demise. For this tragic case, the referral to arrival time was 28 minutes by taxi and the woman was accompanied by a midwife. Fortunately, the woman survived. There was 1 (0.2%) maternal death, which occurred within 24 hours of hospital arrival. She was referred with hypertensive disorder of pregnancy from a non-Platform hospital, but received treatment prior to transfer. She arrived in 132 minutes by ambulance accompanied by a midwife; the fetus died during transport.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Outcomes.

https://doi.org/10.1371/journal.pone.0266932.t005

Discussion

Weak referral systems result in significant delay in the management of obstetric emergencies in many countries [9, 27, 28]. Studies frequently describe factors that lead to referral delay or report problems associated with delay, but few quantitate delay or explore locally driven solutions to address it [27, 28]. This study contributes to the literature in both regards.

This feasibility project utilized WhatsApp to improve communication regarding emergency obstetric referrals in Accra, Ghana and was endorsed by the Greater Accra Regional Director of the Ghana Health Service. Indeed, uptake of the Kybele Referral Platform was rapid; 618 cases were posted within three months and the participant number increased from 69 to 81 during that time. Use has been sustained; at the time of this writing there are 174 individual Platform users. Our working theory of change has been supported. We hypothesized that this low-cost intervention would be readily accepted by providers, and that it could be done in a way that protected patient privacy. The anonymity of the WhatsApp platform made it impossible for us to trace women through the referral process, however, the fact that the platform remains actively used nearly five years later is strong affirmation that the intervention was not only feasible, but helpful to the community of providers in Accra.

In our subsequent analysis of referrals to GARH, it is encouraging that pre-referral treatment was seen more commonly in Platform facilities and referral to arrival times were shortest when urgent cases were transferred by ambulance and when accompanied by a midwife, although attribution of these findings to the Platform cannot be made. The analysis of the Platform transcripts revealed that pre-referral treatment was initiated in 55.2% of posted cases, yet in the referral dataset, pre-referral treatment was provided for only 28.5% of Platform facility patients. It was not mandatory that every patient being referred be posted on the Platform, but when cases were posted and viewed across a spectrum of users, pre-referral treatment may have improved due to higher visibility and scrutiny of the care provided. Alternatively, sicker patients may have been more likely to be posted and therefore more likely to need pre-referral treatment. As has been similarly reported, we found that irrespective of facility category, most high-risk obstetric referrals were unaccompanied by healthcare staff, contrary to national guidelines [2, 6, 29, 30]. When accompanied, we found that the referral to arrival time significantly decreased and this was most pronounced in Platform facilities.

The Platform highlighted failure in the promotion of care along the recommended referral pathway. Platform analysis showed that only 19% of posted cases were received by district hospitals, whereas GARH received 61%. The referral data similarly found that 14% of referrals to GARH came from district hospitals, while 62% came directly from lower-level facilities. Ideally, district-level hospitals should be acting as the intermediary in receiving lower acuity cases, to reduce the high volume of referrals to tertiary hospitals. Bailey and colleagues studied obstetric referral patterns and facility readiness among 977 delivery centers across Ghana [3]. They found that district hospitals accounted for 52% of deliveries and 9% of referrals [3]. Limitations of the Bailey study included no knowledge of where women went once referred or the timeliness of referral, both of which this study addresses. These are important considerations for the ongoing efforts to decongest high-volume referral hospitals.

This work further demonstrates that in the urban setting of Accra, Ghana, significant delay occurs in the obstetric referral process, even among hospitals that were included on the Platform. Research on emergency obstetric referrals in Ghana and other countries is limited, particularly in the quantitative measure of delay in reaching secondary and tertiary levels of care. A study in Mozambique reported that transport and referral delay accounted for 60% of the 712 maternal deaths but details on timeliness were not provided [13]. In Rwanda, neonatal outcomes were significantly worse among mothers whose travel times exceeded 90 min [10]. In the Upper East Region of Ghana, 104 maternity unit officers responded to an oral questionnaire on emergency obstetric interventions and reported a 37 min (range 5–120 min) mean travel time to the nearest referral facility using various means of transport; however, travel time was not measured [4]. Obtaining quantitative data from interviews is subject to recall bias and could be inaccurate. Twenty years ago, Nkyekyer reported on peripartum referrals to KBTH, but timeliness was inadequately assessed [29]. Data were available for 46 (11.6%) of 396 referred patients, and the average transport time to reach KBTH was 78 min. The time from referral decision until facility departure was determined for seven women but was not stated [29]. Our study measured the time interval between referral and arrival at the receiving hospital and provides important detail to a recognized gap in the literature. We found that among the 515 urgent referrals, the median referral decision to arrival time was just under five hours, clearly exceeding the two-hour recommended time interval.

Delay is influenced by more than just the physical distance to a health facility and women in Accra typically do not report “distance” to a health facility as a barrier to care [1]. Many patients represented in the current study arrived from facilities within 20 km of GARH yet still had long referral to arrival times. It is uncertain what happens to women in this time interval. A recent study from Lagos, Nigeria might shed some light [31]. Forty-seven postpartum women residing in the metropolitan area were interviewed regarding timeliness in reaching a hospital from home. Travel times of 5–240 min varied by traffic congestion during the day and security issues and lack of public transport at night [31]. Women reported making stops along the way to pick up support persons and to purchase items deemed necessary for hospitalization. Seventeen of the 47 women were referrals to a CEmOC facility when the initial facility could not provide the necessary care. When women were referred, they were often expected to determine on their own how to reach the referral facility, which prolonged delay and increased cost. The authors report that, despite recognizing danger signs and referral, “pregnant women are faced with conundrums on ‘when’, ‘where’ and ‘how’ to reach CEmOC facilities” [31].

These conditions were likely present in the current study, as we found that 80.2% of referred patients arrived by means of public transport, which did not improve among Platform facilities, and as corroborated by others [2, 3, 6, 18, 30, 32]. This is concerning given that in 2004 Ghana launched a national ambulance service that has been expanded, especially in urban areas [3]. When ambulances are utilized, delay can still occur as patients may need to wait for ambulances located remotely or provide funds to fuel the vehicle [2, 4–6, 29, 30]. However, when traveling by ambulance, the median referral to arrival time was significantly shorter. Nevertheless, the public may perceive that taxis are a faster form of transport than ambulances and this needs to be addressed [33]. In 2002, Nkyekyer found that 65.8% of referred patients utilized taxis or small public buses for transport [29]. In 20 years, little has changed.

In the present study, the leading referral indications were labor dystocia (26.2%), hypertensive disorders of pregnancy (12.3%), and prior uterine scar (11.6%). This is in agreement with other reports, which have similarly found obstructed labor to be the most common referral indication [3, 13, 22, 23, 31]. In Nkyekyer’s 2002 report, only 2% of patients were referred for prior uterine scar [29]; however, this indication has substantially increased [32]. A rising cesarean delivery rate, particularly in Accra, may worsen the existing burden on CEmOC hospitals [1, 30]. A better understanding of labor management practices in lower-level facilities is needed to stem the tide of increasing surgical deliveries [13].

There are several limitations of the current study. We had hoped to gather data on all referred obstetric patients during the September to November 2017 data collection period; however, we captured data for only 597 (47.5%). This was likely due to chart unavailability, data collector unavailability or inattentiveness, and other undetermined factors. Second, we had hoped to gather data on referrals prior to initiating the Platform to allow for a baseline comparison of Platform vs non-Platform facilities, but data collected prior to March 2017 were incomplete and poor quality. Therefore, we could only analyze data collected after the Platform was initiated limiting the attribution of findings. Third, we could not identify individual patients on the Platform for privacy reasons and were therefore unable to determine the arrival times for posted cases. For this reason, specific case outcomes were also unavailable. Fourth, we could have surveyed the referred women to describe their experience, especially in relation to reasons for the delays. For example, we could have queried women as to why they did not report to GARH at time of referral. Indeed, we found some referral to arrival time intervals that spanned several weeks, which is why we focused on urgent referrals. Fifth, delay almost certainly impacted maternal and newborn outcomes, but we were unable to ascertain whether this delay was most significant before leaving the initial site, in route, or both. It would have been useful to gather more information from the source facilities, such as time of departure and reasons responsible for referral delay. Other studies, however, have described these factors which include needing to collect money for treatment or transport, waiting for relatives, needing permission from family members, patient preferences for traditionalist and spiritual practices, first refusal of referral due to fear of disrespect, medical procedures, or surgery, waiting for transport, and traffic congestion [1–3, 5, 6, 13, 27, 28, 30, 34–36]. In some cases, pregnant women are referred from one facility to another on the basis of no bed availability or lack of resources [30, 31, 37]. Fortunately, in the present report, this occurred for only 6 (0.8%) patients.

One advantage to our approach in determining delay from the time of referral decision is that delay likely extends beyond the time required for transport alone. The Platform was helpful in that it provided the receiving hospital advanced warning in order to make preparations to receive a compromised patient. Lack of advanced warning that a patient is coming and omitting a referral slip have been shown to negatively impact preparedness and outcomes at the receiving hospital [2, 3, 6, 34]. Caution in the use of applications such as WhatsApp for healthcare must be exercised. It is paramount to establish guidelines to ensure patient privacy, data security, and phone stewardship [21]. In the current study, participants were forbidden to share photographs or patient identifiers and were reprimanded for posting unrelated content.

It is important to identify and quantify referral-related barriers in the provision of timely CEmOC in low resource countries and this topic remains vastly under-researched [2, 6, 34, 35]. A strong referral system needs communication and coordination among all stakeholders along the referral pathway with tailored referral protocols, cost absorption and efficient transportation to reduce stillbirth, neonatal and maternal deaths [2, 6, 11, 18, 36].

Conclusion

The inability of high-risk women to access CEmOC remains a major challenge in addressing the global burden of maternal and newborn mortality. This study supports the feasibility of using WhatsApp mobile messaging technology to address communication and coordination along the referral pathway while highlighting that referral delays remain problematic. Our analysis indicates that further study and planning is warranted to optimize obstetric referrals in Accra, Ghana. As countries seek to accelerate mortality reduction to meet the Sustainable Development Goal targets, referral systems need further consideration in order to ensure appropriate access, quality and timeliness of care.

Acknowledgments

The authors wish to thank Lynn Harris, Nana Ampofo, and Micheal Asare for assistance with data collection and cleaning and Fiona Bryce and Samantha Tulenko in WhatsApp transcript coding.