The project coordinators defined the review protocol following the PRISMA methodology.17 The review aimed to (1) describe the criteria/algorithms/pathways/models of referral of patients with RMDs from PC to rheumatology and (2) analyse their efficacy (concerning the diagnosis and evolution of these patients) and safety. The following inclusion and exclusion criteria were defined. Studies were selected if they included referral systems for patients with RMDs from PC to rheumatology, regardless of the type of system or underlying disease. No specific comparators were established, and even studies without a comparator were included. If the studies provided data regarding the efficacy and safety of referral systems, we collected data such as early diagnosis or quality of life. Finally, only studies with the following designs were considered: meta-analyses, SRs, randomized controlled trials (RCTs), and observational studies.

With the help of two experienced documentalists, search strategies were generated (Tables S1–S3 of the supplementary material) using Mesh and free-text terms. As a limit, we searched for articles in English or Spanish. The following bibliographic databases were screened: Medline (from inception to July 2021), Embase (from inception to July 2021) and Cochrane Library (from inception to July 2021). All was updated in July 2024. Subsequently, a secondary manual search of the references of the included articles was performed. However, we did not carry out a grey literature search due to the number of retrieved references in the SR. All the references were entered into a reference database (EndNote), to facilitate their management.

Two reviewers independently screened the articles by title and abstract and then by detailed reading. The same reviewers collected data afterwards. In case of discrepancies, they resolved them by discussing in dedicated meetings. To assess the methodological quality of the included studies, we used the AMSTAR 2 for SRs (AMSTAR 2 is not designed to generate an overall ‘score’ but a general rating of a SR taking account flaws in critical domains, which may greatly weaken the confidence that can be placed in a SR),18 the Jadad scale for RCTs (score 0–5; ≥3 is considered good quality)19 and the Oxford quality scale for observational studies (this scale is not designed to generate an overall ‘score’ but a general rating of an study taking account the design and other methodological issues).20 Evidence and results tables were generated.

The survey was designed based on the results of the SR and the experience of the coordinators.21 The objectives were to assess: (1) PCPs level of knowledge and management of RMDs in daily practice; (2) the current referral protocols/systems/criteria across the country (Spain); and (3) the use of digital health tools in PC to guarantee continuity of care and facilitate patients’ referral to rheumatology care. The survey was developed through the online platform Surveymonkey© following a closed and structured questionnaire. It included the following sections and variables: (1) socio-demographic data including age, sex, region of residence, or work experience; (2) level of knowledge and training on RMDs, confidence in the management of this group of diseases and the availability of specific protocols/systems/criteria for referral to rheumatology; (3) use of digital health tools (e.g. electronic medical records, apps) for the management patients in PC, and barriers and facilitators for their use. The survey questions are available in the supplementary material. Given the characteristics of the survey and objectives, a convenience sample of PCPs was selected using a profile grid to assure representativeness (we selected people from different regions, men and women with different levels, etc.). PCPs were obtained from the coordinators’ personal contacts and were then invited to participate in the survey. A pilot study was performed with 4 PC physicians to test whether the questions of the survey will generate useful results. They suggested minor language changes that we incorporated into the survey. The survey then was launched via email on the 30th of October 2021 and closed on the 17th of November 2021. Survey data were downloaded from Surveymonkey© into an Excel sheet. Data analysis was performed using the statistical software Stata 12© (Stata Corporation, College Station, TX, USA). A descriptive analysis was carried out.

A multidisciplinary group of 9 rheumatology experts and 3 PCPs was established. In a modified virtual nominal group meeting guided by a methodologist, they evaluated and discussed the results of the SR and survey. Based on these data and their experience, they proposed a set of criteria and other considerations for the referral of patients with (suspected) RMDs from PC to rheumatology. National and local characteristics of the health system and usual PC practices were also considered. After several alliteration processes, the definitive referral criteria and other aspects were generated to be included in the app (future development).

The search strategies yielded 2152 citations (Fig. 1). After eliminating duplicates and the screening processes, 42 articles were excluded (Table S4 of supplementary material), and 32 (5 with the secondary search) were finally included. Table 1 summarises the main findings of the SR (see Tables S5 and S6 of the supplementary material).The quality of the included studies was generally moderate or low and were conducted in countries with different health systems and organisations. Most of the selected articles were cross-sectional studies related to diagnostic validation of the referral criteria defined by the rheumatologists.10–12,22–25

Figure 1.

Flow diagram of systematic review studies.

On the other hand, referral criteria are mainly focused on a specific RMD. The most frequent ones are RA10–12,22–26 and spondyloarthritis (SpA).11,23,26–38 We found significant variability in the criteria even for the same disease. There were no referral criteria for all of the RMDs, and criteria for purposes different from diagnosis (e.g., risk management) were scarce.11,28 Only one article generated urgent referral criteria.13

The number of items included in the referral criteria varied, ranging from 326 to 18.11 Likewise, the types of items differed widely. Some were clinical (e.g., related to the anamnesis and examination), others to laboratory parameters or imaging techniques. At least two studies evaluated the use of screening systems of varying complexity.13,39

Very few studies evaluated the effectiveness of the referral criteria. The study outcomes included the time saved to the disease diagnosis,26,33 disability improvements24,30,33 or the duration of the first visit.33

A total of 37 PC physicians from all over the country participated in the survey, of whom 69% were women, 72% reported more than 20 years of practical experience, and almost 92% worked in PC general practices.

We found that 70.6% of the participants had additional rheumatology training (more than what is usually done in the training of medical specialists). It is worth noting that 66.6% did so through online training courses, 62.5% by reading scientific journals and 45% by attending congresses. However, the average level of safety when diagnosing and treating patients with immune-mediated rheumatic diseases such as RA, lupus or SpA, using a scale from 1 (very unsure) to 10 (very sure), was 5.79±1.53. However, concerning patients with musculoskeletal diseases such as osteoarthritis or osteoporosis, the mean is 7.85±1.43.

Regarding the availability of a specific protocol or referral system at the health center for patients with RMDs, 39.4% of the participants commented that they have one (Table 2 shows some examples).

Finally, 74% of respondents considered using digital health tools to help guarantee continuity of care and facilitate the referral of patients from PC to specialized care. 9.7% said they were unclear, and 3.2% said they were useless.

On the other hand, only 9.7% use some apps to refer patients to specialized care. The most used are the SER-SEMFYC (for autoimmune diseases) and QxAAPP (for antithrombotic treatment).

Table 3 depicts the characteristics considered necessary for apps to be helpful in routine clinical practice for referring or managing patients. According to the participants in the survey, the most important features were that PA participated in their development (84%), that they are based on scientific evidence (77%), that the items to be assessed are simple and accessible (77%), that they are free of charge (74%) and that they also have information explaining the pathology, evidence and explanations of the algorithm or decision tree (74%).

In the modified virtual nominal group meeting, 12 rheumatologists and PCPs (42% women from eight different Spanish regions with more than 20 years of clinical practice) discussed and proposed the referral criteria (see Table 4 and supplementary material). They are grouped into seven categories based on symptoms/signs, diseases, or treatment complications. Beginning with the primary option, additional ones are gradually established to specify the situations in which it is advisable to consider a referral. The reasons for referral are based on diagnostic suspicions, the possibility of requiring specific diagnostic techniques performed in rheumatology, like musculoskeletal ultrasound or synovial fluid analysis, therapeutic techniques also performed in rheumatology, such as infiltration, and the main complications of most specific drugs in the speciality. For example, a referral is recommended for patients on classical or biological disease-modifying anti-rheumatic drugs and high levels of transaminases.

Some are straightforward, direct criteria, such as assessing a referral in patients with low-back pain >3 months in patients aged <45 years if the pain is inflammatory. Others, however, require a multi-step assessment of symptoms and signs, as in the case of systemic immune-mediated (see Table 4).

Along with the criteria, the experts provided definitions/explanations to increase their reliability, such as validated definitions of arthritis or inflammatory low-back pain.

In addition, each criterion incorporates the suspected diagnosis(s) or complication(s) of an already diagnosed RMD. For example, in the case of monoarthritis, the following differential diagnosis is proposed: crystal-induced arthritis (gout, chondrocalcinosis), septic arthritis, and osteoarthritis.

In case a referral is recommended, a series of complementary tests to facilitate the work of the rheumatologist are suggested. For patients with suspected systemic immune-mediated diseases like lupus, this proposal incorporates antinuclear antibodies, complement levels, urinalysis, etc.

The referral criteria also include and describe red flags in each of the 7 referral criteria that indicate an urgent referral to the rheumatologist. For example, some of the red flags for patients with inflammatory low-back pain would be the presence of fever without other explanation (that might suggest an infection at that level), neurological deficits, night pain that do not respond to medication, etc.