Designing and describing an electronic referral system to facilitate direct hospital admissions

In France, hospital emergency rooms have been in crisis for many years. The model conceived in the 1960s was centred on the management of life-threatening emergencies and serious traumas. However, these activities have changed substantially over the years, and emergency rooms have become a source of chronic overcrowding [1]. The same observation has been made at the international level [2, 3]. Part of this overcrowding is related to inappropriate emergency department (ED) visits. The lack of standardization of this concept makes the estimates of published rates not very comparable [4]. It has been estimated that between 13.5% and 27.4% of the ED visits that occurred in 2013 in France were ‘nonurgent’ [5] and that 16.1% of visits in England between 2015–2017 fell under the same category [6]. Many studies have been conducted to identify the causes of these inappropriate ED visits. Most of these studies have focused on patient characteristics (social vulnerability, need for free care, anxiety, consumerist attitude, etc.) [7] or on the inadequacy of the primary care offer (lack of preventive activities, medical deserts) [4, 7, 8].

The reorganization of the system is becoming a public health priority. One area for improvement is to facilitate direct admissions, which are defined as hospital admissions that do not go through the ED [9]. However, this direct admission process requires good coordination between primary care providers (PCPs) and hospital specialists to ensure the fluidity of the primary-secondary care pathway. Few studies have addressed the impact of the lack of primary-secondary care coordination on the progressive use of the ED by PCPs as a gateway to hospitalization for nonurgent patients [9].

In France, Derame et al. showed that 10% of patients hospitalized via an ED visit could have benefited from direct admission [10]. The authors concluded their study by remarking on the need to set up telephone hotline systems between PCPs and hospital specialists to facilitate direct admission. The investigation conducted by this team in 2008 showed that despite the implementation of such a system, direct admission was still not scheduled often enough. More than one patient out of three was still hospitalized in downstream wards via an ED visit during working hours even though they were eligible for direct admission [11]. Dijon et al. experimented with a telephone hotline for the direct admission of geriatric patients, which showed a reduction in the length of hospitalization in the group that benefitted from direct admission and a reduction in the time to rehospitalization [12]. However, the use of a telephone hotline poses a certain number of problems described by Hermant, namely, difficulty in reaching the hospital specialist, time spent scheduling hospitalization and the occasional loss of information about the request, which results in the absence of progress in the management of the patient [13].

The use of digital tools can help overcome these difficulties. Various electronic referral systems have been tested in North America and have been shown to allow general practitioners to make nonurgent requests to specialists [14]. These systems have been proven to be effective by reducing the time needed to access specialists and reducing the costs associated with hospitalizations [14, 15]. These systems have thus enabled quantitative and qualitative improvements in communication between PCPs and hospital doctors. However, they were designed for requests for specialist advice or consultations but not for direct hospital admission.

The objective of our study was to describe the development and implementation of an electronic referral system in an area of the Paris suburbs (Hauts-de-Seine Sud—France) aiming to facilitate direct hospital admissions.

The methods were performed in accordance with the relevant French guidelines and regulations. The survey of PCPs was declared to our hospital Data Protection Officer (DPO) and recorded in the data protection register of Paris-Saclay University Hospital. The medical data were collected as part of the SIPILINK patient medical management. Consequently, this study did not fall within the framework of French regulations on research involving human persons, and written patient consent was not required. A patient’s nonobjection was collected by the PCP (an informational note detailed the purpose and processing of the data) before the data were entered. As the data were processed by the hospital staff managing this care, this study did not require additional authorisation from an IRB or from the Ethical and Scientific Committee for Research, Studies and Evaluations in the field of Health (paragraph 2 of article 65 of French law no. 78–17 of 6 January 1978: https://​www.​cnil.​fr/​fr/​la-loi-informatique-et-libertes#article65).

A primary-secondary care coordination system to facilitate direct admission was implemented with the aim of increasing the rate of direct admissions to the hospital. The system was based on a digital tool, namely, SIPILINK, which was designed to address coordination difficulties. Communication difficulties were one of the causes reported in our survey of PCPs within the study area.

The participation rate for the PCP survey was low (18.5%). Low participation of general practitioners in surveys is a problem frequently described in the literature [24]. Explanatory factors with material and immaterial influences have been studied, and the participation rate has been shown to have important variability due to the context of each country, varying between 6% in Germany and Austria and 90% in Malta (median participation rate: 30%) [24]. We did not find any article in the scientific literature specifically studying the participation rate of general practitioners in declarative surveys in France, but by consulting several thesis works on equivalent subjects, we found various participation rates ranging from 5% [25] to 37% in the DEMOMED 75 study [26]. There is no current consensus on a method to be adopted to improve these participation rates. It seems that electronic survey methods often have lower response rates than paper mailed surveys [27]. Therefore, we favoured face-to-face contact to propose the survey. Despite this methodological limitation, the results obtained on our sample were in line with results published in the literature and reinforced our interest in developing an innovative coordination system. Indeed, the PCPs stated in our survey that they were satisfied with their collaboration with hospital physicians, particularly with the quality of their exchanges. This is consistent with the results of Marshall et al. [24]. However, the survey found that the frequent use of referral to the ED as a point of entry to hospitalization was due to the difficulty of contacting a hospital practitioner to schedule a direct admission.

One of the pillars of the system that was put in place was to allow a system of exchanges that optimized the medical time on both the hospital and primary care sides. On the primary care side, the interest was in avoiding the waiting time on the phone for PCPs to access a specialist who is able to respond to their request. On the hospital side, the interest in the system was in preventing specialists from interrupting their hospital activities with telephone calls, thus risking answering too quickly without having all the elements necessary to make the best decision. Indeed, the system allows the hospital to access the requests when they are available to respond to them under good conditions, possibly after accessing the patient's hospital medical file to specify the response and to start the care planning. Finally, the system allows for the traceability of exchanges as well as the sharing of these exchanges among all of the actors involved in the care (PCPs, hospital physicians, care pathway managers).

The description of the response times expressed by the PCPs showed that requests for a response "as soon as possible" were not frequent (17.7%); rather, the most frequently requested timeframe was within 72 h (37.1%), followed by a timeframe within 24 h (33.9%). The results concerning the real response times carried out by the hospital showed that the system largely met the doctors' expectations in terms of response time since only 10.5% of these responses were made after 24 h. This responsiveness is an extremely important point for the project’s sustainability. This result was possible due to the implementation of cross-disciplinary care pathway managers who monitored and reminded the hospital physicians in cases of nonresponse. In e-referral schemes evaluated by Tuot et al., 91–95% of the responses were made within 72 h, and the median time for an electronic response from a specialist was 24 h in the Los Angeles Safety Net Program eConsult System [28].

The SIPILINK system was particularly useful during the first peak of the SARS-CoV-2 pandemic in France, with a massive enrolment of PCPs in March and April 2020. The choice of technology (web-based application) as well as the development carried out by the hospital's IT team allowed for a high level of responsiveness when the pandemic started. It was indeed possible to create a new section almost instantaneously to address the needs related to COVID-19. This organizational innovation thus enabled PCPs to feel less isolated during this period of health crisis and to facilitate access to specialized advice and hospital care during a particularly tense period in terms of access to hospital care. Moreover, the satisfaction of PCPs with this type of system has been found in the international literature [29].

The development of this system also presented certain difficulties. On the hospital side, the compartmentalization of project management between hospital administration and doctors made it difficult to provide on-the-ground projects and did not facilitate the implementation of this project [30]. Thus, the first few months of implementation were relatively slow due to obstacles at both the hospital and primary care levels. Hospital staff are currently overworked and lack the time and availability to participate in innovations [31]. PCPs are sometimes entrenched in their practice habits and reluctant to change. These observations are not specific to the French health care system; similar observations have been noted internationally. In a review of the literature, Osman et al. described the most commonly mentioned obstacles to the use of e-consultation systems by general practitioners, namely, the need to modify their behaviour, the modification of their workload, and the loss of immediate contact with their usual specialist referrers or, in contrast, contact with specialists with whom they are not used to collaborating [29]. Our system differs in its purpose from the electronic addressing systems developed in the USA and Canada [18] because these systems are essentially aimed at referral to hospital specialist consultations, whereas our system is essentially aimed at referral for direct hospital admission. The impact of interventions on direct admission and their potential effects on reducing the rate of inappropriate emergency room visits has rarely been studied in the literature.

Moreover, few studies have examined the impact of the lack of coordination between the PCP and the hospital physician on inappropriate emergency room visits or on the rate of direct admission [9], thereby showing that this dimension, which we seek to address with our system, has not yet been the subject of much research. Andronikof et al. showed in a study published in 2008 that it was possible to organize scheduled hospitalization in only 40% of cases where hospitalization was indicated. In 11% of the cases, the solution proposed by the hospital contact person was to refer the patient to the emergency room [32]. As Leyenaar et al. pointed out, direct admission to the hospital is not suitable for all clinical situations [33]. It is indeed more appropriate for patients who require urgent care or a rapid imaging examination [9]. This is why we have excluded the management of urgent situations in the SIPILINK users’ agreement. The urgent nature of care is thus left to the discretion of the PCP. The reasons for requests in our study showed that the reasons were very rarely related to an emergency evaluation, with the exception of requests related to COVID-19. A before-and-after analysis of the direct admission rate after the implementation of the system could not be carried out because of interference with the SARS-CoV-2 health crisis. Indeed, the care pathway for admitted patients was modified in spring 2020, with systematic passage to the ED for admitted patients to test them for COVID-19 diagnosis. COVID-19 patients at the height of the spring 2020 epidemic wave constituted almost all of the patients who were hospitalized in the study hospital; furthermore, since the epidemic is still ongoing, this parameter is currently not assessable.

The difficulties of primary-secondary care coordination are most often addressed in the literature in the hospital primary care context [9, 34, 35] and essentially from the communication aspect, particularly in relation to hospital discharge. For example, Ekwegh et al. developed a model discharge letter (structured letter) to improve this communication [34]. In the primary care hospital context, delays in access to a specialist consultation are most often studied, with delays estimated at between 6 and 12 months. The estimated consequences are the loss of chance linked to a delayed diagnosis, the repetition of diagnostic examinations, the cost to the health system [36], and inappropriate visits to EDs [28]. Various e-addressing systems for specialist consultations have been tested in the USA and Canada [14, 17, 18, 37, 38] and more recently in England [39].

The first experimentation found in the literature was conducted at San Francisco General Hospital in 2005 [38]. A web portal was developed by this hospital to set up a primary-secondary care e-referral system. The system is very similar to the one we set up with the SIPILINK system. Each department has a designed medical referent who reviews and responds to requests from PCPs. He or she can then quickly schedule a consultation or hospitalization or ask for additional information or investigations before deciding on the patient's referral. These exchanges are then integrated into the hospital's electronic medical record [40]. This process is relatively comparable to that implemented with SIPILINK. In Canada, Liddy et al. described the development of a platform called "The Champlain BASE" (Building Access to Specialist through E-consultation), which is an application that allows primary care physicians to submit a request for nonurgent specialty care [14]. This platform was set up in 2010, and more than 10,000 requests were submitted within the first 5 years of operation. The authors were able to show that the platform reduced the time required to schedule a specialist consultation and, in particular, that 3.4% of the requests for advice for which the PCP did not initially consider a consultation resulted in a consultation. Similarly, a platform called e-Consult was developed in 2012 in the Los Angeles area to facilitate access to specialist consultations by filtering requests from PCPs via an e-form [28]. The e-form is then read by a doctor of the requested specialty who decides, according to the elements provided, whether to refer the patient to a consultation for the specialty requested. This system has seen a rapid increase in the number of requests made by PCPs and has reduced the time it takes to gain access to specialist doctors. The median response time is 24 h, and 25% of e-consultations are resolved without a physical consultation.

The reason for the lack of work on the impact of primary-secondary care coordination on the rate of direct admission is probably linked to the fact that most of the work on primary-secondary coordination is carried out by PCPs who have pointed out the consequence of this lack of coordination on primary care management, particularly after hospital discharge. Few hospital doctors have taken an interest in primary-secondary care coordination to analyse the difficulties generated on the hospital side. In summary, it seems that the subject of primary-secondary coordination has not yet been sufficiently decompartmentalized between the two health care sectors to improve the bottlenecks in the entire process [41].

The sustainability of the project is based on 3 factors: communication, territorial approach, and incentives. The start-up of the system was indeed slow but was boosted by the COVID-19 crisis. As the first wave of COVID-19 started a few weeks after the launch of the system, we were only able to do very little communication. In the context of a health crisis during which professionals are overwhelmed with information, we therefore favoured word of mouth among professionals. This process was relatively effective. Nine months after the launch, without any particular communication, 36.6% of the PCPs installed in the area were registered. To maintain the link with users and bring new PCPs on board, we will send a biannual newsletter to all PCPs in our territory, with the objective of reporting on the use and evolution of the service. This electronic distribution will be ensured by the general practitioner trainers’ network of the Paris Saclay University and by the Osmose territorial health network, whose role is to promote local initiatives by health professionals.

Another lever of sustainability is to extend the use of this primary-secondary care coordination system to other hospitals in the territory. It is indeed important that PCPs do not have to use different types of electronic communication tools depending on the hospital to which they wish to refer a patient. Currently, one of the challenges physicians face is the increasing number of digital solutions that are aimed at addressing certain sequences of the care pathway. This fragmented multiplication of tools will require knowledge about and the manipulation of numerous interfaces that may hinder the proper management of the patient’s care pathway. A model such as that developed in San Francisco can be adapted to the organization of our local health system [38]. The integration of the system into the shared electronic medical record currently under construction [42] will be one of the elements that determines its generalization. The third lever of sustainability of the project will also require the consideration of the remuneration of the actors involved in this coordination. Indeed, as doctors emphasized in a study in Catalonia, organizational factors such as coordination time are just as important as digital tools [43], which means that funding must be provided. Finally, Ekwegh et al. insisted on the importance of collaborative work between PCPs and hospital doctors for the development of their tool, a collaborative tool that is quite similar to the tool that we have developed in our project [34]; however, there are still very few formalized primary-secondary care networks that aim to develop collaboration and perpetuate this type of initiative [44]. Thus, it is necessary to develop them.

The initiative presented in this publication has been supported by local stakeholders, which is a key element of its success. The first months of its implementation have shown its value in reducing the isolation of PCPs in their practice (particularly during the first wave of COVID-19) and in better organizing hospital responses to primary care requests to encourage direct admissions. To ensure its sustainability, the next step will be to extend its use to other hospitals in the region, which, in the current context of an abundance of digital offerings for PCPs, is the next challenge of the project. The deployment of digital tools for primary-secondary care coordination must be done within the framework of a territorial strategy for the health care organization defined by local actors and supported by the health authorities.