Referral reasons of type 2 diabetes patients from general practitioners to diabetes specialists: a cross-sectional observational study

Abstract

Background

Diabetes is a global public health concern. The follow-up of patients with diabetes is often undertaken by general practitioners (GPs), but referral rates to specialists show variations. The primary objective of this study was to describe the factors associated with the referral of type 2 diabetes patients from GPs to diabetes specialists.

Methods

An observational survey-based cross-sectional study was conducted with GPs practicing in France between 17 May 2023 and 13 July 2023. Physician characteristics, referral rate and attitudes towards referral were collected. Characteristics associated with frequent referrals were evaluated using a multivariable logistic regression model.

Results

A total of 325 GPs located in 52 departments (mean age 43 years) responded to the questionnaire. Most responding GPs were women (63%). Most GPs (76%) stated that they rarely or never referred their patients with type 2 diabetes to a diabetes specialist. The most frequent barrier to referral was delays in accessing a specialist (57%), and it was often cited both in the infrequent referral group (56%) and the frequent referral group (58%). In multivariable analysis, higher referral rates were associated with physician age (Odds Ratio OR per year 1.04, 95% Confidence Interval CI 1.01 to 1.07), diabetes care network membership (OR 2.81, 95% CI 1.15 to 6.88), referrals motivated by the introduction of insulin therapy (OR 2.73, 95%CI 1.44 to 5.34) or to consolidate communication about therapeutics and compliance (OR 2.34, 95%CI 1.18 to 4.67), expecting advice regarding new medication such as SGLT-2 inhibitors (OR 2.08, 95% CI 1.11 to 3.98), and the mention of patient refusal as one of the main barriers to referral (OR 1.85, 95% CI 1.03 to 3.38) were associated with higher referral rates. Conversely, doubts about the added value of the diabetes specialist was associated with infrequent referrals (OR 0.25, 95% CI 0.08 to 0.66).

Conclusions

These factors indicate a possible role for lasting professional relationships, where trust built over shared network memberships and years of practice can facilitate referrals. Newly established GPs may find it beneficial to contact specialists in their region to discuss referral procedures. This could facilitate the referral of patients and improve access to specialist care.

Background

Diabetes is a global public health concern. In 2021, the International Diabetes Federation estimated that 536.6 million people aged 20 to 79 were living with diabetes worldwide [1]. The prevalence of diabetes is estimated to be 4.9% in France, with type 2 diabetes accounting for 93% of cases (more than 3 million of people) [2]. The follow-up of type 2 diabetes patients often relies on general practitioners (GPs), diabetologists, or a combination of both. A study conducted in 66 GP surgeries located in South Wales (United Kingdom) showed a referral rate of 4.2% (range: from 1.2 to 16.3%) which is consistent with the hypothesis of a follow-up mainly undertaken by GPs [3].

Cooperation between GPs and diabetes specialists is vital. Data extracted from the electronic records of 17,493 patients showed that therapeutic inertia, defined as a lack of modification or intensification of therapy when a situation should require it, was encountered in 18% of type 2 diabetes cases [4]. A meta-analysis found a 1.4% improvement in taking hemoglobin A1C tests when a GP and specialist communicated face-to-face, by telephone or by videoconference [5]. Comparing patient outcomes between diabetic patients treated by diabetes specialists and patients treated by GPs is difficult, since the former often have more severe diseases than the latter. The outcomes for achieving target blood pressure were similar in a Japanese study [6]. Nonetheless, specialists often need to manage diabetes in frail or complex patients, therefore collaboration between both types of physicians may be beneficial [7]. The average confidence level of GPs in treating diabetes ranges from 50 to 85% [3], and GPs with lower confidence have a higher referral rate [3].

Current French guidelines recommend to visit a specialist only in specific situations such as blood glucose levels ≥ 16.7 mmol/L, hemoglobin A1C levels ≥ 10%, pregnancy, need for insulin therapy, or complications [8]. As a chronic condition with substantial impact on the patients’ lives, diabetes is one of the conditions covered by the French “universal health protection” policy, which means that diabetic patients do not incur out-of-pocket expenses for consultations and essential treatments. However, depending on the data source, 20–30% of cases may take over three months for a referral to a specialist to occur [9].

Referrals to specialists may be necessary to consolidate therapeutic communication with patients that are anxious, suicidal, non-compliant to treatment regimens, or to raise their awareness about the seriousness of the condition. However, barriers to cooperation still exist for healthcare professionals (sometimes stemming from disagreements regarding treatment preferences and goals) as well as for the patients (who may not feel the need to include a new actor in their list of healthcare providers) [10]. GPs often prefer to refer patients to specialists practicing in hospitals, which are not always available nearby in rural areas. Moreover, referral rates vary by country [11] which may indicate a potential need to explore the sources of these differences. A qualitative study including 12 GPs, 12 diabetologists and 24 patients has shown that GPs and diabetologists see themselves as competing rather than complementary physicians. Although the study had identified some of the determinants of referral, including reasons for referral and obstacles to referral, it was conducted in a single department on a small sample of respondents [10]. Therefore, the objective of this study was to identify GP characteristics associated with the frequency of referral of type 2 diabetes patients to a diabetes specialist.

Methods

Study design

An observational cross-sectional study was conducted with French GPs practicing in mainland France using an online questionnaire. The questionnaire was sent between May 17, and July 13, 2023 via three channels:

• the mailing lists of medical professional unions (Unions Régionales des Professions de Santé Médecins Libéraux, and Unions Régionales des Médecins Libéraux),

• the mailing lists of local antennas of the French National Order of Physicians.

- Social networks used by physicians to hire replacement physicians in case of absence.

Since being part of the French Order of Physicians is compulsory in order to practice medicine in France, use of these sources ensured that most GPs could be represented in the sample. Every known local instance of the aforementioned organisations was contacted with the request to distribute the questionnaire to all GPs. These sources were free to send reminders if they considered that it was needed. Therefore, the target and source population for the study was all practicing GPs in France.

Non-practicing GPs or those practicing outside of France were excluded from this study. The items in the questionnaire were either 10-point Likert scale or multiple-choice questions relating to demographic characteristics and the type of medical practice. Questions using the Likert scale were used to assess the perceived ease GPs felt in referring patients and their satisfaction with the disease management provided by diabetes specialists (Supplementary Table 1).

Primary outcome

The primary outcome was the referral rate reported by physicians. Referral rates were grouped into two classes: frequent referrals or absent/rare referrals. Frequent referrals corresponded to more than one in ten patients with type 2 diabetes over the past year. This number reflected the 10.6% of annual diabetes specialist consultations recorded in the ENTRED study [12], and was consistent with referral rates observed in other European countries [3].

Data collection

The GPs’ sex, age, length of time in practice, regional department of France, urban or rural location of practice, as well as the travel time from the location of the practice to the nearest specialist were recorded. Travel time was considered to be long if it exceeded a threshold of 30 min, in normal circulation conditions. GPs were also asked whether they were meeting pay-for-performance targets for quality of care, according to a list of indicators aligned with public health objectives. The pay-for-performance (Rémunération sur Objectifs de Santé Publique: ROSP) scores considered were the scores specifically pertaining to diabetes care. For example, one of the targets for diabetic patients undergoing treatment is that ≥ 92% of patients have at least two HbA1c measurements per year. If the indicator was not reached, it was considered to be below the target value. If a GP can attain this objective, the GP gets a financial bonus on the basis of public health achievements. Each GP’s number of patients and their number of patients with chronic diseases were also recorded. The percentage of patients with chronic diseases was calculated for each physician as the ratio of these variables (using the number of patients with chronic diseases as the numerator). Using the postal code of the area in which they practiced in order to provide an indicator for each geographic region, the social deprivation factor (French Deprivation index: FDEP) [13] was used to analyse data according to socio-economic status differences of the areas corresponding to the postal code of the GPs’ practices (the indicator is based on mean education levels, unemployment rates and average income of each area).

Other collected variables regarding the type of practice included: single practice (practices with only one GP), group practice, practice in a multi-professional health center known as the Maison de Santé Pluriprofessionnelle (MSP), or in an accredited MSP receiving funding from a regional health agency. The three last categories were grouped together during the analysis to facilitate the interpretation of results and to obtain a more parsimonious model at the multivariable analysis stage. Membership of a territorial professional healthcare community known as the Communauté Professionnelle Territoriale de Santé (CPTS), or of a healthcare network for diabetes management, were also recorded. The location of the practice was also recorded. To achieve an acceptable sample size in each category, regions were grouped as northern and southern regions. The departments considered to be in Northern France were: Normandie, Bretagne, Hauts-de-France, Grand Est, Pays de la Loire, Centre-Val de Loire, Ile-de-France and Bourgogne-Franche-Comté. Conversely, Occitanie, Auvergne-Rhone-Alpes, Provence-Alpes-Cote d’Azur, and Nouvelle-Aquitaine were considered to be in Southern France.

We asked GPs whether they worked with certified nurses trained in therapeutic education. In France, an association known as Action de Santé Libérale en Equipe is dedicated to the provision of training and support for these nurses, so that they can cooperate with GPs to provide chronic patients with follow-up care and therapeutic education. We also asked if they worked with Advanced Practice Nurses (APNs). This role is conferred to selected nurses who complete a specific two-year degree and have at least three years of experience. The reasons for referral, as well as the benefits of, and barriers to the referral of patients with type 2 diabetes were collected based on possible factors found in the literature [10, 14] (see Supplementary Table 1 for a complete list).

Statistical analysis

Continuous variables were expressed as means and standard deviations (SD), or medians and quartiles Q1–Q3 if the variable was asymmetrically distributed. Categorical variables were expressed as numbers and percentages (%). Referral frequency was converted into a binary variable to facilitate interpretation. Physicians who stated that they “never” or “rarely” (only one in 10 patients in the past year, no patients or only one patient was referred) referred patients were grouped together in the “Infrequent referrals” group. Physicians who answered that they often referred patients (more than one in 10 patients in the past year) constituted the “Frequent referrals” group. Chi-squared tests (or Fisher’s exact tests, as appropriate for smaller samples) and Student’s t tests (or Mann-Whitney U tests in case of asymmetric distribution) were performed to compare the groups according to their referral rate. Reasons for referral, benefits expected of referral and barriers to referral were compared between the infrequent referral group and the frequent referral group. They were also compared according to the type of practice (individual or group practice), to practice location (urban or rural), to diabetes healthcare network membership, and to CPTS membership.

To facilitate the interpretation of results, subcategories representing similar concepts were grouped in a single category for some of the variables of the initial questionnaire. All types of grouped practice, all travel times exceeding 30 min, all pay-for-performance scores below the main target value were grouped to be interpreted as single categories.

A multivariable logistic regression model was performed, predicting the probability of being a physician that frequently refers diabetic patients, with adjustment on variables significant at the p ≤ 0.10 threshold in bivariate analysis, with additional a priori manual selection on clinical relevance. In the event of multicollinearity, only one variable from the group of colinear variables was selected for inclusion in the model, depending on the relevance of the variable according to expert opinion. The analysis of clinical relevance was realized by two physicians (JC and SS). Regarding pay-for-performance indicators, only attainment of the main target was included in the bivariate analysis and multivariable analysis. The model’s performance was assessed using the Area Under the Curve (AUC) [15]. Analyses were performed using R software, version 4.3.3 (R Foundation for Statistical Computing, Vienna, Austria). A p-value < 0.05 was considered statistically significant.

Ethical considerations

In accordance with law N° 2021 − 300 (March 5, 2012), this study (a questionnaire-based study with physicians, without reliance on individual patient data) did not require approval by an ethics committee. This study was not subject to the Jardé law since it did not fall within the scope of Article R.1121-1 of the French Public Health Code.

Results

Responder characteristics

A total of 325 GPs located in 52 administrative departments across France participated to the study, with a mean age of 43 ± 11 (SD). Most participating GPs were women (63%). Most of the respondents (246 out of 325, 76%) stated that they rarely or never referred their patients with type 2 diabetes to a diabetes specialist. Most GPs (87%) worked in a group practice or in multi-professional health centers. One hundred and thirty-six GPs (42%) were members of a CPTS and 33 (10%) were members of a diabetes care network (Table 1).

Table 1 Characteristics of 325 French general practitioners according to frequency of referrals to diabetes specialists

GPs who frequently referred patients were generally older (with a mean of 42 ± 11 vs. 45 ± 11 years in the infrequent referrals and frequent referrals group, respectively; p = 0.04). GPs in the frequent referral group were located in urban settings more often than GPs in the infrequent referrals group (42% vs. 56% for the frequent referrals group, p = 0.04). Membership of a diabetes care network was also more often seen in the frequent referral group (7.7% vs. 18%, p < 0.01). GPs who made frequent referrals found it easier to access a diabetes specialist (mean score: 4.8/10 vs. 5.6/10, p = 0.01).

Reasons for referral, benefits expected of referral, and perceived barriers

Overall the most frequent motive mentioned by responders for referring a patient with type 2 diabetes was diabetes imbalance (77%) (Table 2). Over half of the referrals (54%) were for the initiation of insulin therapy. One tenth (10%) of GPs said they referred their patients with regards to the diagnosis of diabetes.

Table 2 Reasons, benefits and barriers to referral provided by general practitioners according to referral frequency

Regarding the benefits of referral, 63% of GPs stated they were seeking to gain a fresh perspective on patient management and 47% resorted to referral in order to obtain an opinion on new medication (e.g. gliflozins) (Table 2). There were no statistically significant differences in reasons for referral, benefits expected from referral, and barriers to referral according to the location of practice, to the type of practice (single practice or group practice), to participation to a diabetes care network or to CPTS membership (Supplementary Tables 2, 3, 4 and 5).

Regarding the main barriers to referral, 57% of GPs felt that the wait time for an appointment with a diabetes specialist was excessive. Patient refusal was cited by 45% of physicians. Distance was cited as one of the main barriers predominantly in rural areas (2.7% vs. 23% for urban and rural physicians, respectively; p < 0.001), although the overall percentage of positive responses for this barrier was relatively low (14%; see Supplementary Table 2). There were no significant differences in obstacles reported by GPs according to whether they practiced solo or in one of the grouped types of practices, whether they were members of a diabetes care network, or according to membership of a CPTS (Supplementary Tables 3, 4 and 5).

In bivariate analysis, the main reasons for referral that differentiated the “infrequent referral” and “frequent referral” groups were willingness to refer in order to start insulin therapy (mentioned by 48% and 71% in the infrequent referral and frequent referral groups, respectively, p < 0.001) and diabetes imbalance (74% and 87% respectively, p = 0.01) (Table 2). The main expected benefit from referral that differed between the two groups was getting an informed opinion before the introduction of innovative treatments such as SGLT-2 inhibitors (Gliflozins), which was more often cited in the frequent referral group (42% vs. 63%, p = 0.001). The main barriers to referral that differed between the two groups were doubts about the added value of the specialist (this answer was less often seen in frequent referral rate physicians: 26% vs. 6.3%, p < 0.001), and refusal by the patient, which conversely was more often cited by patients with frequent referrals (40% vs. 58%, p = 0.01).

Multivariable analysis

In multivariable analysis (Table 3), age (OR per year 1.04, 95% Confidence Interval CI 1.01 to 1.07, p = 0.005) was a strong predictor of referral frequency. An increase of 15 years of age would therefore increase the adjusted probability of belonging to the frequent referral group by 12% (95% CI 3.4 to 22) for a typical GP in the sample (the 15 years period is provided as a representative example). Membership of a diabetes care network (OR = 2.81, 95% CI 1.15 to 6.88) was also associated with increased referral frequency. This variable accounted for a 15% (95% CI 2.5 to 28) adjusted increase in the probability of belonging to the frequent referrals group (average marginal effect, corresponding to the adjusted effect across the sample). Willingness to refer a patient for onset of insulin therapy (OR = 2.73, 95% CI 1.44 to 5.34), and wishing to bolster communication about therapeutics and adherence to treatment (OR = 2.34, 95% CI 1.18 to 4.67, p = 0.01) were also associated with increased odds of referral. Concerning the benefits of referrals, GPs that expected advice on new medication referred patients more frequently (OR = 2.08, 95% CI 1.11 to 3.98, p = 0.02). Doubts about the added value of diabetes specialists was the strongest barrier to referral in multivariable analysis, with an OR of 0.25 (95% CI 0.08 to 0.66, p = 0.01), corresponding to an adjusted decrease of 20.3% in the probability of belonging to the frequent referral group (95% CI -4.4 to -35.2) (Table 3). Conversely, mentioning patient refusal as a barrier to referral was associated with a higher referral rate (OR = 1.85, 95% CI 1.03 to 3.38, p = 0.04). The model had good predictive performance, with an Area Under the Curve (AUC) of 0.778 (Table 3).

Table 3 Odds ratios of characteristics associated with frequent vs. infrequent referrals to diabetes specialists in France

Discussion

In this nationwide study conducted in France, we found that age, membership of a diabetes care network, referrals to start insulin therapy or consolidation of communication regarding therapeutics and compliance, and expecting advice on new medications (e.g. SGLT-2 inhibitors) as one of the benefits of referral were independent predictors of frequent referrals to a diabetes specialist. Citing patient refusal as a barrier was also associated with a higher referral rate. Conversely, the main barrier associated with infrequent referrals was doubt about the added value of diabetes specialist management. These associations persisted after adjusting for practice location. Therefore, the physicians who often referred patients were more likely to seek the opinion of a specialist before initiating therapeutic decisions (be it insulin therapy or treatments involving new molecular classes). The univariate analysis showed that more referrals were made by GPs who practiced in urban settings. This factor was close to statistical significance in multivariable analysis (p = 0.08). However, doubts about the value of referral on the GP’s part (OR = 0.25; 95% CI 0.08 to 0.66, p = 0.01) were more likely to prevent referrals than the setting of the GPs practice.

Studies conducted in other settings suggest that the difficulty of accessing specialists in rural still remains a problem (as reported by 23% of responders in a household survey conducted in rural Saskatchewan, Canada) [16]. A literature review by Perron et al. (2022) assessed inter-professional cooperation in rural areas in Canada. Peer groups and continued medical education through courses, workshops and simulations were found to be among the initiatives already being carried out to enhance cooperation between physicians. Suggestions for improvements were formulated, stressing the benefits of working in small teams, and the role of financial compensation [17].

Diabetes care networks are organized along these lines, with patient management based on cooperation between the GP, APN and the paramedical staff. This reduces the impact of patients with chronic conditions on the GP’s patient base. Participating in multidisciplinary cooperation to alleviate the physicians’ workload may prove to be worthwhile, as treatment and prevention targets, for example retinopathy screening rates, are often unmet in European countries [18].

More than three quarters of participating GPs reported that they would refer their patients to specialists in case of diabetes imbalance, and more than half indicated that they would refer patients to initiate insulin therapy. In the multivariable analysis, GPs willing to refer patients for the initiation of insulin therapy also had higher adjusted odds of referral. These categories correspond to referral reasons mentioned in the French guidelines for the treatment of diabetes, although referral for complications appeared less frequent, possibly because complications occur less often than the former reasons. Diabetes imbalance was the main reason for referral to specialist clinics for GPs based in the United Kingdom [19]. It was also the main motive in a study conducted in a North American military hospital [20]. While access to a multidisciplinary follow-up appeared as a consensual reason for referral in a Canadian study using qualitative methods [21], it was not highly prevalent in the answers provided by GPs in our study. Although a study has indicated that some GPs may be uncertain when confronted with the diagnosis of diabetes [22], diagnosis was a minor reason for referral in our study.

No differences were found in reasons for referral according to the type of practice. Overall, the most frequently cited barrier to referral across the sample (regardless of the high or low referral rate of physicians) was the wait time before an appointment (which affected more than half of the physicians, similarly in both groups). Out-of-pocket expenses were described as one of the main barriers to referral by 18% of GPs. Although in France the expenses of diabetic patients are taken charge of by the government, patients with socioeconomic deprivation could still struggle to obtain specialized care, due to delays in the refunding of some expenses, and differences in consultation delays between the public and private sectors. These patients could therefore be reluctant to accept being referred to a specialist. Such differences are difficult to capture in studies with a using the place of residence as a proxy for socioeconomic status, which probably explains why the difference was not statistically significant for this variable in our study. Poor communication between the parties was also found to limit referral opportunities in the literature [23].

Regarding the limitations of our study, as the study was based on a questionnaire pertaining to past events, it may have been difficult for GPs to remember the exact number or proportion of patients referred. Although frequent referral was defined by using one out of ten patients as a threshold, the question could have been interpreted in a subjective manner and the answers could be prone to recall bias. Less physicians than expected indicated that they often referred patients to diabetologists, which could be a consequence of a subjective interpretation of the question. Referrals for pregnancy, and for specific types of diabetic imbalance were not evaluated. Fewer referrals do not necessarily imply poorer quality of care. Some GPs may be able to carry out patient management and communication with other specialists themselves. It would be interesting to study the relevance of patient referrals, but the factors used to assess this relevance could be difficult to establish.

The mean age of responding physicians was younger than the average age of practicing physicians during the year in which the study took place (53 for men and 48 for women) [24], possibly due to the mode of recruitment of the study, which partly relied on a call for participation on social media. This means that the conclusions of the study may not apply in populations of older physicians. Although respondents were often from newer generations of physicians, willingness to participate in the study was also unevenly distributed across regions and this may have induced a form of selection bias.

There was no difference in patient referrals according to CPTS membership, or to practice with a nurse trained in therapeutic education, despite a good representation of these two factors in our study (41.8% and 29.8% of respondents, respectively). GPs working with APNs were in the minority in our study, but this profession is reasonably new. Diabetes specialists could have a different role according to the availability of nurse professionals with advanced roles, including diabetes specialist nurses [25]. The role of each actor should be well defined, as the multiplication of actors does not always lead to improved outcomes [26]. Pay-for-performance scores were not associated with frequent referrals in our study. Some of these indicators are self-declared by the physician, which may question their reliability. However, medical services paid for by national solidarity are routinely verified by the national health insurance system. Moreover, as only 48% of physicians have reached or exceeded the target value, this did not seem to be a major problem in our sample.

The development and standardization of care networks in the management of diabetes appear to represent interesting prospects for change. Their actions are diverse, but they enable cooperation and coordination of care between general practitioners, diabetologists, nurses, psychologists and dieticians. They can facilitate follow-up examinations, therapeutic education and access to sports. Exchanges between staff members take place on a regular basis by mail, telephone or via secure Internet access. As actors within the networks could share best clinical practices, participation in the networks could possibly decrease the heterogeneity observed in the care of diabetic patients across European countries [27]. However, care networks are often sectored according to former administrative regions or departments, physicians may not have equal opportunities of joining.

Telemedicine may be a solution worth developing in the future in order to improve access to specialists for patients in rural areas [28]. Occasional consultation days in rural areas conducted by specialist physicians, as is already being done in some multi-professional health centers, may be worth exploring.

Conclusions

The association of age with higher referral rates suggests that referral rate could be driven by positive social ties with surrounding specialists, combined with knowledge of their actions and their added value for the treatment of patients. Newly established GP may find it beneficial to contact diabetes specialists within the region in order to discuss a referral procedure. This measure could have a positive impact on the care provided to diabetic patients, especially in regions showing demographic changes in the medical population.