As shown in Figure 1, the literature search generated 2,167 potentially eligible publications and further reference screening yielded 7 additional publications. After duplicate removal, 129 full texts were assessed for eligibility. Finally, 30 articles were included in this review. The results are presented as 22 unique projects, which include multiple articles for most projects.
Figure 1: Preferred Reporting Items for Systematic Reviews and Meta-Analyses” (PRISMA) flow diagram of the study selection process (18)
Figure 2 summarizes the review`s results. It visualizes the challenges that influenced the design of interventions, as well as the main components of collaboration and the role distributions identified in the literature.
Figure 2: Overview of challenges in healthcare provision, components of the interface of collaboration and role distributions as identified from the literature.
HCPs: healthcare professionals; GP: General practitioner.
1: Patient- and disease-related factors refer to challenges in healthcare provision that stem from patient characteristics or characteristics of their conditions. These factors influence the collaboration between specialists and general practitioners, but cannot be modified by the collaboration.
2: Contextual factors originate both from the healthcare system and the collaboration at the primary and secondary care interface, i.e., these factors influence the collaboration between specialists and general practitioners, but can also be modified by this collaboration.
Project characteristics
Rationale for project
Most authors designed their intervention with either, challenges in the healthcare system or characteristics of a specific patient population in mind. One important reason was fragmented care, which was a barrier to care provision. Reasons for fragmented care were a lack of communication and cooperation in care that was vulnerable to miss patients at risk. It ultimately resulted in the delayed provision of care (19, 20). The lack of unclear roles of involved HCPs, guidelines and referral processes limited the patients’ access to care (21-24). Lack of knowledge and awareness of GPs on the condition resulted in a shortcoming of care and preventive measures (25-27). Authors specified that highly prevalent health conditions were focused on in the interventions. Authors reported on the respective condition’s burden on healthcare facilities or system resources such as high consultation rates (23, 24, 28) or costs (29-34). Furthermore, projects presented that the health condition of interest is a leading cause of morbidity, hospitalizations and death (30-33, 35-40). The current organization of care and its coordination was often considered suboptimal or even outdated for the specific health condition. Finally, the conditions’ complexity was mentioned. Interestingly, only few projects reported on what they defined as complex patient care or complex needs even though most of them used this term to describe patient-related characteristics. These definitions included: polypharmacy (23, 24), a specific amount of follow-ups (29), “stage- and symptom-dependent care that comprises multimodal interventions” (26, 27), presence of comorbidities (34, 41), presence of multiple practitioners (42) and patients’ “ongoing management of their [condition] is beyond the capability of their usual GP” (43, 44).
Origin and study design
Of the 22 original projects, five were located in the United States (19, 23, 24, 31-33, 35-37, 45, 46), four in Canada (21, 22, 29, 30, 42); two projects, respectively, were located in Australia (43, 44, 47), Brazil (41), Spain (28, 48), Switzerland (26, 27, 38) and the Netherlands (20, 49). One project was based in Germany (25), Italy (34) and New Zealand (39).
Eight projects applied a before-and-after design with outcome measurements pre and post intervention (19, 23, 24, 30-34, 41, 45), whereof three before-and-after designs further compared the intervention group to a control group (38, 47, 48). Additionally, three projects compared the intervention group to a control group, which was usual care (28, 42-44). Five projects were RCTs (20, 26, 27, 35-37, 39, 49). One project used an array of methods for their evaluation; the authors applied a randomized stepped-wedge design, a before-and-after and a controlled design separately (21, 22). Furthermore, two protocols for RCTs (29, 46) and one protocol for a before-and-after design (25) were included in this review.
Seven projects applied randomization, and of those, two projects randomize on a patient level (21, 22, 35-37) and five on a healthcare provider level (e.g. long-term care home, GP practice) (20, 26, 27, 29, 39, 49). Three projects blinded the involved healthcare professionals (HCPs) with regard to the allocation sequence (21, 22, 35-37, 39) one mentioned specifically that blinding was not possible (21, 22) and three gave no information whether any study participants were blinded (20, 26, 27, 46, 49).
Twelve projects reported an intervention period of 12 months (20-22, 26, 27, 29, 30, 38, 42-47, 49) and one project of 10 months (23, 24). Nine projects evaluated the intervention for more than one year with a mean duration of approximately 40 months, ranging from 15 to 156 months (19, 25, 28, 31-34, 39, 41, 48).
Patient population
Projects involved various patient populations with either specific or unspecific diagnoses. Specific diagnoses included kidney diseases in one (42) and specifically at a chronic stage (chronic kidney disease (CKD)) in five projects (19, 20, 30-34). Furthermore, three projects included patients diagnosed with chronic obstructive pulmonary disorder (COPD) (26, 27, 38, 45), whereas a combination of asthma and COPD (41), chronic hepatitis B (CHB) (47), chronic heart failure (CHF) (35-37) and coronary artery disease (CAD) (46) were investigated in one project. A diagnosis of diabetes type 2 was focused on in two projects (43, 44, 49). One project defined the patient population of interest as a mix of several specific diagnoses, where at least one of the following was present: diabetes, cardiovascular disease, COPD, asthma or cardiovascular risk factors (21, 22). Six projects characterized their patient population independent of diagnoses such as long-term (29, 39), end of life care (25) or care for patients with multiple chronic conditions (23, 24). Furthermore, one project focused on heart diseases in general (48) and another on gastroenterology and hepatology without specific diagnoses (28).
Organization of care and components of interface
Interventions implemented process and system changes to enhance collaboration between primary and secondary care. The measures focused on education and knowledge transfer, as well as shifting responsibilities to additional HCPs.
Education and knowledge transfer
Education and training for GPs and primary care staff was widely applied. Two of the included projects solely designed continuous medical education (CME) programs as the intervention. Adams et al. created a CME program that targeted primary care providers and aimed to improve their knowledge/comprehension, self-confidence, and, ultimately, clinical practice (45). Uniquely, Pang et al. developed a mentor-mentee program with individual training sessions, regular regional learning meetings and case consultations. The educational aspect of the intervention focused on early detection and identification of patients with CKD (30). A Dutch project focused on educating primary care staff in “blood pressure management and treatment, cholesterol lowering, blood-glucose management, and lifestyle advice” for patients with CKD (19, 23, 24, 28, 29, 49). Fortin et al. arranged a chronic disease prevention program with a focus on patients’ self-management. Included HCPs received training in motivational interviewing to help reduce the burden of chronic diseases and related high-risk lifestyles (21, 22).
Most projects combined educational programs with other components in the intervention. Tinetti et al. encouraged physicians to review selected patient cases together and to identify patient priorities besides offering CME. The physicians were stimulated to align their care towards patient priorities and the same outcomes (23, 24). Health condition specific education was part of Martins’ et al. study to improve the physicians’ knowledge before inviting a medical specialist to perform joint consultations in the GP practice (34, 41). Before opening specialized outpatient clinics, Santoro et al. provided teaching. Especially, early identification of the at-risk population, effective prevention and treatment strategies and inappropriate drug utilization were taught to GPs. Within the outpatient facilities GPs were then asked to profile patients, assess their co-morbidities and renal insufficiency stage, plan lab tests and referrals. Furthermore, if appropriate, GPs developed personalized therapy and nutrition plans (34, 41).
Care arrangements between GPs and specialists were usually based on knowledge transfer and hands-on teaching without planning formal education events (19, 23, 24, 28, 29, 49). As an example, Falces et al., integrated a cardiologist into a primary care practice to support follow-up (48). Similarly, the project of Hynes and Porter introduced new HCPs in the patient-centered medical home, who shared their information in multidisciplinary team meetings (31-33, 38). In an Italian study, co-management of patients was encouraged and the specialist was involved more, as soon as the GP needed assistance in the management (34). Several interventions provided screening tools (19), guidelines (19, 25-27), treatment protocols (47), standardized discharge processes (46) and defined referral procedures (19, 26-28, 42, 47, 48). Usually, these standards were developed before the project and in accordance with project participants and experts to ensure applicability.
Electronic consultation tools were mainly introduced to provide remote assistance for decision-making and support by specialists for GPs. Ong et al. established a co-management arrangement which was virtual and supported with guiding referral processes that visualized patients’ pathways (42). Other electronic platforms were incorporated to store patients’ health records (20), exchange documents (42), and facilitate direct interaction (29, 46). Ho et al. focused on providing telemedicine consultation for GPs and provided educational documents on evidence-based care of geriatric patients (29, 46). Reasons for integrating e-consultation tools included improved specialist access and reductions in unnecessary in-person referrals (29, 42). Furthermore, they added diagnostic and therapeutic support as well as a way to provide continuing health education (46).
Shifting responsibilities to other HCPs
Additional HCPs besides physicians were widely integrated at the primary-secondary interface. Bekelman et al., Connolly et al. and Porter et al. integrated nurse coordinators or specialized nurses to arrange multidisciplinary team meetings (39). Additionally, Porter et al. added a community health worker and a pharmacist to the team in a long-term care home (31-33) and Bekelman et al. included a social worker to perform psychosocial assessments (35-37).
Their roles varied between projects concerning their involvement in the patients’ care processes. Nurse practitioners, nurse educators, specialized nurses, nurse coordinators and practice nurses were the professionals most included in interventions. Their level of involvement varied, but nurses engaged rather actively and performed patient assessments, developed and reviewed care plans, coached staff, took over patient education and shared responsibilities with GPs (19, 21, 22, 30-33, 35-39, 43, 44, 47, 49). Other HCPs with active roles included pharmacists, physiotherapists and community health workers who provided patient education on profession-specific care (31-33, 38). In a Canadian project, the research staff was engaged actively and performed pharmacological consultations and medication reviews (29). Rather passive involvement in the care provision included medical education of nurses, physician assistants, medical students, respiratory physiotherapists without any further specified role (26-28, 45, 46).
Role distributions
Some projects defined roles prior to the intervention and others openly gave physicians the task to define them themselves. The following synthesizes the roles, how they were distributed among GPs and specialists or shared between them.
Many interventions assigned clearly distinct roles to GPs and specialists. The more severe the disease and symptoms, the more likely a specialist was involved in patient care. On the other hand, the GP was often expected to support with problems of little to moderate severity. Accordingly, the patients were referred to the specialist for specialized assessments and treatments (23-27, 34, 47). Furthermore, several projects expected specialists to provide medical education to GPs. GPs were then expected to enhance their knowledge and services by receiving CME (23, 24, 26, 27, 30, 43, 44). Pang et al. assigned GPs to specialists, acting as mentors. In brief, an assigned mentor was responsible for regular education sessions and offered one-on-one consultations, but little supervision during daily practice. In this project, the GPs knowledge on diagnosing and managing patients with CKD was evaluated (30). Another project encouraged a hospital to distribute specialized patient care plans to the GPs upon referral. Management was transferred to the GPs but overseen with an electronic treatment protocol by the hospital team. The primary care team was evaluated regarding their ability to treat and monitor patients with CKD (47). In a Swiss project, the GP was asked to follow a checklist closely and their adherence was monitored by the research team (26, 27).
Other studies included interventions on joint consultations or patient case discussions. They allowed for more direct interaction between physicians (20, 29, 41, 42, 46). For instance, van Gelder et al. integrated a web-based consultation tool to enable GPs to align their care with nephrologists by consulting them. The GP was responsible for patient care and data entry into patients’ electronic medical records. Specialists had access to the same information and provided streamlined advice (20). Falces et al. integrated a specialist into the primary care team as a new member to assist in patients’ follow-up and to distribute the patients between care levels. The physicians were able to establish their own referral criteria and actively discuss their different responsibilities. In this case, specialists provided medical advice when needed, and had weekly consultation sessions with GPs (48). Even though the specialist took the lead in joint consultations with the GP and patients in a Spanish study, the GP was still the main person of contact for the patient (28). In the project of Fortin et al., GPs were in charge of long-term follow-up of patients, while the specialist took a rather passive role and had regular contact with the GP for support accordingly (21, 22). Haley et al. motivated GPs and specialists to develop tools addressing specific barriers and root causes for problems related to collaboration. In this project, the GP and specialist were considered equally important for patient care and co-managed within the scope of their expertise. The GP was in charge of patient management and sought specialized support when needed (19).
In projects that integrated further HCPs into primary care institutions, the GP was expected to supervise and to provide support. The GP was usually in charge and responsible for patient care but delegated certain tasks and responsibilities (31-33, 35-38, 43, 44, 49). Compared to a traditional primary care setting, a project taking place in a long-term care home is co-managed by the GP and other HCPs without the GP taking the lead or being superordinate (39).
Outcomes and results of projects
The majority of projects analyzed patient-related outcomes. Overall, the results were positive and reported improvements in outcomes. Several projects reported on improvements in patients’ health condition-specific outcomes. They included HbA1c level (43, 44), blood pressure (49) or estimated glomerular filtration rate (20, 34). Falces et al. declared that an improvement in body mass index, better control of blood pressure and cholesterol were achieved (48). No change was reported in the primary outcome, a patient-reported disease-specific HRqoL questionnaire, but secondary outcomes (depression, fatigue and anxiety) improved (35-37). Two projects evaluated referral rates to the specialist and found a decrease at project termination (28, 41). Connolly’s et al. measured a decrease in avoidable emergency department admissions (39).
Self-reported physician outcomes included improved clinician confidence (45), satisfaction (48), increased knowledge (30, 45), heightened attention to communication and awareness for a health condition (19). Furthermore, the participation rate of GPs in attending one-on-one case discussions with specialists was reported (30). Additionally, process-related outcomes included improved documentation of outcomes by the GP (19) and an improved disease-specific medication dispensing (41). Finally, some interventions were reported to increase costs slightly (20) or to not increase resources (48).