Discussion

Our experience demonstrated that EHUs can be a safe and effective model of care that has a number of key strengths. By co-locating EHUs in existing medical facilities, these were able to detect new chains of transmission by screening walk-in patients who had a non-specific febrile illness and were seeking general health services. Screening and early isolation of these patients protected both HCWs and patients without EVD from nosocomial transmission on the general wards. This enabled the health facilities to remain open for non-EVD emergency care for the duration of the protracted outbreak, whereas many other health facilities without EHUs had to be closed down. For example, before the construction of the EHU at Rokupa Government Hospital, the hospital had 11 HCW infections in 3 weeks. After construction of the EHU at this hospital, there were none.

Suspected patients who tested negative for EVD but had another serious illness (such as appendicitis or life-threatening complications of HIV) could immediately access further inpatient or outpatient care within the same facility. Our experience was that the EHUs were safe for staff and patients, provided reliable diagnoses and initiated early bundles of care covering a variety of likely pathologies (P Arkel, D Youkee, CS Brown, et al. Quantifying the risk of nosocomial infection within Ebola Holding Units: a retrospective cohort study of negative patients discharged from five Ebola Holding Units in Western Area, Sierra Leone. Paper submitted to Tropical Medicine & International Health; March 2016).

The EHU model has system-wide benefits. Using existing facilities meant that EHUs could be deployed early in the outbreak, and rapidly constructed.20 Start-up times were rapid due to the benefits of utilising existing logistics, waste management facilities and staffing schedules. It was much less expensive to adapt existing structures than to build new sites. Together these reduced resourcing requirements, such as capital expenditure or international or national health workers, which freed up resources to be deployed elsewhere. The use of existing supply chain and other government systems reduced the need for parallel structures that were expensive and challenging to establish, and risked undermining local long-term systems. Leadership of the EHUs by the MOHS and local HCWs could be maintained, thereby building up local institutional capacity and helping to ensure local ownership and community acceptance. Integration within the MOHS allows the EHUs to coordinate effectively with other aspects of outbreak response, thus strengthening the health system alongside the EVD response.

EHUs also demonstrated flexibility and could be adapted as the epidemic grew and receded. In EVD endemic countries, such as Uganda and the Democratic Republic of the Congo, sporadic re-emergence is an ongoing concern. Capacity must remain to isolate, test and treat EVD: this must be responsive, scalable and adaptive to local epidemiology. The EHU model works well within a resilience framework.13 These can be diverse; in Freetown, they were located at primary and secondary care facilities as well as specialist centres for obstetrics, paediatrics and psychiatry. They can be multifunctional and adaptive, allowing for the isolation and screening of other potential outbreaks such as cholera, Marburg or Lassa fever; it might not be feasible to manage diseases transmitted through the airborne route, such as Middle East Respiratory syndrome (MERS), as different infection control methods are required. They can easily be reactivated to enable EVD care with trained, on-site staff if and when the need should arise. The hub and spoke model, using Connaught as a base, allowed for resilience and staff support and training from a central site.

Our experience also showed that the EHU model has important limitations. The EHUs usually only provided isolation, diagnosis, and early treatment but then had to refer confirmed cases to ETUs for further care. Where no ETU beds were available, patients sometimes had to travel for up to 6 hours by ambulance to remote ETU facilities, which most likely had a detrimental impact on their outcomes. The medical care available to patients, while meeting national and international standards, was relatively limited; the EHUs did not have the capacity to provide all patients with intravenous fluids or have access to novel therapeutics that were being trialled. These challenges were not unique to EHUs, but there were some factors about this model that made certain procedures more difficult or risky for staff: utilising existing buildings limited the ability to optimise layout and infection control processes; and EHUs operated with much lower staffing numbers compared to other types of facility, as well as relying largely on government health workers who were often less well paid or highly qualified than international counterparts. Higher staff to patient ratios and more senior supervision for staff would enable higher standards of care to be provided, and indeed this was achieved at some EHUs later in the outbreak.

The EHUs were most useful in an urban setting, where there was easy direct patient access to a secondary or tertiary care facility: they were able to pick up and detect new cases that presented to hospital in a setting where contact tracing was very challenging; they could utilise existing buildings in areas where there was little or no open ground for tented facilities; and could be established early in an outbreak and be kept open for long periods with minimal resources, which was important as new chains of transmission can rapidly be brought into urban hubs from rural areas with little notice. While EHUs are still relevant for hospitals in rural areas, community-based models of care may have a more significant role in this context. The function of EHUs was largely restricted to medical interventions, with limited involvement in community sensitisation or contact tracing. The use of pre-existing infrastructure sometimes meant that optimal layout was difficult to achieve and compromise was often necessary, such as having to locate incinerators or screening points at sites distant from the isolation beds. However, sites did often have appropriate physical locations, for example by being adjacent to entrances, for ease of entry into the EHU from screening areas.

The use of existing systems, such as medical supply chain and risk allowance, had benefits; however, this could also add pressure to an overstretched system and create challenges. Managing this required effective collaboration with government partners, flexibility (eg, switching to different types of PPE, when necessary), and proactive troubleshooting when problems arose. Ultimately this worked well but we were unable to eliminate the need for parallel systems completely; for example, we needed a supplementary supply chain for some items, and there were still challenges. These included staff strikes due to lack of hazard pay. Similarly, the partnership approach involved more stakeholders (local hospital, supporting partner, MOHS, donor) and this could make decision-making slower and more challenging. Managing this required trust, good communication, mutual empathy and strong, and sustained leadership from all sides.

Overall we believe the EHU model achieved the four goals set out in the introduction above, although given the chaotic, multidimensional and rapidly evolving nature of the outbreak and the lack of good data it is difficult to prove this or unequivocally show a causal relationship. The EHUs will have significantly reduced the number of cases in the community, given that 1159 EVD cases were isolated in EHUs from the community in the Western Area at a time when no other isolation beds were available.21 The nosocomial transmission rates in EHUs have been shown to be extremely low and were certainly lower than would have been achieved in a general hospital setting. We believe that EHUs in Sierra Leone made an important contribution to reducing EVD transmission. There is no data on survival rates of patients with EVD who were isolated in EHUs compared to those remaining in the community or presenting to other forms of facility; however, since all suspected patients were immediately started on treatments (such as oral rehydration and antibiotics) that are known to reduce EVD mortality, and since these were largely unavailable in the community and there were no available beds elsewhere, we can infer that EHUs reduced mortality for patients with EVD. Some, but not all, general health services were maintained in hospitals with EHUs; the Connaught Hospital Emergency Department remained open and general medicine wards remained largely at capacity during the outbreak, although other services were significantly reduced, such as elective surgeries which were reduced by 97%. The extent to which general health services were maintained varied between departments and across different hospitals, and varied over time. However, hospitals with EHUs were more likely to remain open. For instance, the Ola during Children's Hospital (the equivalent paediatric hospital to Connaught Hospital in Freetown) closed completely for 2 months from September to November 2014 until an EHU was established and the hospital was reopened. There were similar closures of hospitals in Monrovia, Liberia. The reduction of health worker mortality is a complex issue: many hospitals without an effective EHU saw high levels of health worker infections and these infections reduced significantly when an EHU was opened (as demonstrated above at Rokupa Hospital). However, the health worker infection rate was not eliminated completely in hospitals with an EHU; since it is likely that the hospitals would have closed without an EHU and staff would have remained at home, it could be argued that EHUs kept staff at risk. However, many health worker infections occurred when health workers provided informal care in the community (which would likely have increased if hospitals had closed) and without EHUs, the overall scale of the outbreak would likely have been greater (putting everyone in the community, including health workers, at increased risk). We believe that our data supports the claim that EHUs largely achieved their goals.

We learned a number of important lessons from our involvement in the response to the Ebola outbreak in Sierra Leone. An effective and resilient health system is the cornerstone of any response to a health crisis and it is essential to invest in getting the basics right when times are good, before the emergency. We were too slow to challenge some of the conventional wisdom from previous outbreaks, particularly the belief that EVD would not cause major urban outbreaks. The Government of Sierra Leone and its partners ultimately did shift into emergency mode, but we could have moved from a ‘development’ mindset to a ‘humanitarian’ one more quickly, spending greater resources more rapidly, reducing bureaucracy, suspending non-critical activities and adopting a different relationship and form of communication with partners. We demonstrated that government services, including the public health system, are not only capable of playing a safe and central role in a health crisis and humanitarian emergency but are the fastest and most efficient way of doing so. We believe that supporting government health facilities as the primary responders to a crisis should be the starting place of any international relief efforts through close collaboration. Crises often become protracted and are likely to recur in future. However, funding for any response soon starts to dry up once the height of the emergency is over; it is, therefore, advisable to plan for the long term even from the beginning, such as building permanent facilities or training local health workers. Finally, good command and control can enable a far more effective use of existing resources. For example, maximising the use of existing bed capacity is more effective than building more beds, but this requires up-to-date operational data and clear protocols for decision-making.