General Practitioners' Propensity to Cooperate With Other Health Professionals in the Management of Patients With Multimorbidity and Polypharmacy: A Cross-sectional Study.

Background. Cooperation between general practitioners (GPs) and other health professionals appears to help reduce the risk of adverse events linked to polypharmacy for patients with multimorbidity. We investigated the existence of different GP proles according to their opinions and behaviors about such cooperation and studied the association between these proles and the GPs’ characteristics and deprescribing behaviors. Methods. Between May and July 2016, we performed a cross-sectional survey in a panel of French GPsabout their management of patients with multimorbidity and polypharmacy, focusing specically on their opinions of healthcare professionals’ roles and interprofessional cooperation. We used an agglomerative hierarchical cluster analysis to identify GP proles and then multivariable logistic regression models to study their associations with these doctors' characteristics and deprescribing behaviors. Results. We identied four proles of GPs according to their cooperation propensities: GPs from the “intensive” prole (14%) were favorable to cooperating with various health professionals, including delegating some prescribing tasks to pharmacists; GPs from the "moderate" prole (47%) had favorable opinions about health professionals’ roles, except for this specic task delegation; GPs from the "selective" prole (27%) tended to work only with physicians; GPs from the "low cooperation" prole (12%) didn’t appeared interested in cooperation. These proles were associated with different professional characteristics. Conclusions. Current health policies encourage interprofessional cooperation for the management of patients with multimorbidity. Our study provides information for understanding disparities among GPs regarding working with other professionals who deal with their patients and suggests possible ways to improve cooperation.


Background
Multimorbidity, de ned as the coexistence of two or more diseases in the same individual [1], leads to polypharmacy, which is associated with several iatrogenic risks -including drug-drug interactions and adverse drug events -and premature mortality [2,3]. Multimorbidity also increases the use of primary and secondary healthcare services [4,5], thus complicating the coordination of care by general practitioners (GPs) [6,7]. The number of specialists involved in the management of patients with multimorbidity and their prescriptions also increases the risk of adverse drug events [3]. GPs must regularly review the prescriptions of patients with multimorbidity and polypharmacy and may need to deprescribe medications with an unfavorable risk-bene t balance. Interprofessional management of polypharmacy involving case managers, nurse practitioners, or pharmacists, can decrease potentially inappropriate prescriptions [8], especially when it involves prescription review by pharmacists [9,10]. Effective cooperation between GPs and other health professionals may also improve these patients' functional status, professionals' adherence to recommended practices, and the quality of GPs' practices (e.g., monitoring of patients with diabetes, or benzodiazepine prescriptions following guidelines) [11].
Accordingly, health policies in several countries encourage interprofessional cooperation and multiprofessional practices (e.g., multidisciplinary or multiprofessional group practices and healthcare networks) [12][13][14]. Several qualitative studies have addressed these professionals' experiences and perceptions of such cooperation [7,15,16]. But little is known about how and the extent to which GPs cooperate with other healthcare professionals (medical specialists and non-physician professionals as pharmacists, nurses, physical therapists etc) and in managing patients with multimorbidity and polypharmacy, and whether their propensity to cooperate may in uence their own practices.
This study's objectives included the: 1) identi cation of different pro les of GPs, according to their opinions and attitudes towards cooperation with other healthcare professionals when managing this patient population; 2) study of the personal and professional characteristics of GPs associated with these pro les; and 3) the assessment of whether such pro les are associated with a propensity for deprescribing.

Design and population
We conducted a cross-sectional survey about the management of patients with multimorbidity and polypharmacy, nested in a panel of GPs in private practice in France. The panel design has been described in previous publications [17,18]. In brief, we randomly selected GPs from the exhaustive French database of health professionals (Répertoire Partagé des Professionnels de Santé) between December 2013 and March 2014. The sampling was strati ed for age, sex, annual number of consultations and house calls, and medical density of each GP's municipality of practice. We excluded GPs with fewer than 5 patient visits weekly, those planning to retire within 6 months, and those practicing exclusively alternative medicine (e.g., acupuncture or homeopathy). At inclusion, participants answered a short questionnaire about their professional characteristics. The subsequent interview later included questions about their practice organization: solo practice (GP practicing alone in a single o ce), group practice (several physicians practicing together), and multiprofessional practice (GPs working with several health professionals including nurses, physical therapists, psychologists, etc).

Data collection procedure and questionnaire
Professional interviewers used a computer-assisted telephone interview system to collect data between May and July 2016 with a standardized questionnaire. It was developed with a multidisciplinary group of experts, based on a literature review and the results of two qualitative focus groups, one of four and the other of ve GPs. We pilot-tested the questionnaire among 50 GPs to clarify and validate the questions.
We de ned multimorbidity as the presence of multiple chronic diseases in a single person. The questionnaire addressed GPs' opinions about their role and that of medical specialists (e.g. cardiologists, endocrinologists etc.) in the management of prescriptions in patients with multimorbidity and their opinions about their cooperation with pharmacists regarding polypharmacy (4-point Likert scale from strongly disagree to strongly agree, see Additional File). We asked GPs about the frequency of contacts with specialists or pharmacists for medication management (4-point Likert-like scale: never, sometimes, often, and very often), and the usefulness of other non-physician health professionals' participation in the management of these patients (Yes/No). We also asked GPs about their propensity to initiate the deprescription of medications they consider inappropriate (4-point Likert-like scale: never, sometimes, often, and very often). A "don't know" answer was also proposed for each question.

Statistical analysis
Data were weighted to match the national GP population according to the strati cation variables. For objective 1, multiple correspondence analysis (MCA) was used as a pre-processing step for further cluster analysis: this allows to reduce the dimensionality of the data and to transform categorical variables into continuous ones (factor coordinates). Then results we conducted an agglomerative hierarchical cluster analysis (AHCA) to identify cooperation clusters (pro les) of GPs, according to their opinions and behaviors regarding cooperation with other health professionals (Additional le contains the items included in the analysis). We used the minimum inertia lost to identify the optimal number of clusters; this corresponds to a minimal within-cluster variance (individuals with maximum similarity in each cluster) and a maximal inter-cluster variance. Some items used for MCA and AHCA analyses had categories accounting for less than 10% of the answers. These categories were then regrouped with the closest category (e.g., often/very often). The "don't know" answers were excluded from the analyses.
For objective 2, we used the cooperation clusters as dependent variables and conducted multinomial logistic regressions to study their associations with GPs' personal and professional characteristics (age, sex, practice organization and participation in a continuing medical education (CME) course). Logistic regressions were adjusted for workload, medical density and reported proportion of patients with multimorbidity on GP's list.
Finally, for objective 3, we studied whether the cooperation clusters (the variable of interest) were associated with GPs' propensity to deprescribe medications they considered inappropriate (dependent variable). We used a multivariable logistic regression model, adjusted for GPs' strati cation variables (age, sex, workload, and the medical density in their practice area).
All analyses were based on two-sided P values, with statistical signi cance de ned by P ≤ .05; they were conducted with SAS 9.4 statistical software (SAS Institute, Cary, NC).

Results
Of the 3724 eligible GPs, 1712 (46.0%) agreed to join the panel in 2014, of which 1266 (73.9%) were still participating at the time of this survey in 2016. Among them, 1183 (93.4%) completed the multimorbidity/polypharmacy questionnaire. Table 1 describes the participants' characteristics. GPs' pro les regarding interprofessional cooperation The cluster analysis identi ed four pro les of GPs, based on their opinions about different professionals' roles regarding the management of patients with multimorbidity and polypharmacy and their behaviors and opinions about cooperation with these professionals ( Table 2). GPs from the rst pro le (14% of the sample, intensive cooperation pro le) had favorable opinions about the roles of both specialists and other healthcare professionals in the management of this patient population and about their cooperation with them. Only this group had a majority (71%) agreeing with the delegation to pharmacists of medication management (prescription review and modi cation). GPs from the second pro le (47%, moderate cooperation) had more positive opinions than average about pharmacists' knowledge and role in managing polypharmacy, but they were nonetheless mainly opposed to letting pharmacists modify patients' prescriptions. They were more favorable than average to cooperation with various health professionals for the management of these patients, including delegation of consultations to nurses, but had more negative opinions than average about specialists' role in this management. GPs from the third pro le (27%, selective cooperation pro le) were likely to cooperate with other physicians and had more positive views than average about specialists' roles in medication management. They tended to oppose the delegation of prescriptions to nurses or pharmacists and had unfavorable opinions about the role and cooperation of pharmacists. GPs from the last pro le (12%, low cooperation pro le) expressed the most frequent negative opinions of cooperation, task delegation, and other professionals' roles in the management of polypharmacy. Associations between GPs' pro les, their characteristics, and deprescribing propensity Multinomial logistic regression (Table 3) with the low cooperation pro le as the reference, showed that GPs with a moderate cooperation pro le were the youngest. The latter, together with those with an intensive cooperation pro le, were also more likely to have participated in CME course. GPs' pro les were not associated with their workload, their medical density, or the self-reported proportions of patients with multimorbidity on their lists. They were not associated with practicing in multiprofessional groups, but those with an intensive cooperation pro le worked less frequently in group practices. Finally, GPs' propensity to deprescribe inappropriate medications was lower in the moderate and selective cooperation pro les than in the low cooperation pro le, but the association was not strong (Table 4).

Discussion
We found various pro les of GPs' propensity to cooperate and share the management of prescriptions for patients with multimorbidity and polypharmacy with specialists and other health professionals. While most GPs recognized pharmacists' knowledge about medications and their adverse reactions, only a minority would agree to share prescription management with them.
GPs belonging to cooperative pro les (moderate and intensive) pro les had taken more CME courses than those in the low cooperation pro le. GPs' propensity to initiate the deprescription of inappropriate medications was weakly correlated with their propensity to cooperate.

Comparison with literature
Qualitative studies have shown that some GPs feel that the involvement of many specialists in the management of these complicated patients results in the fragmentation of care. Combined with a lack of communication between physicians, this may lead to increase harm from polypharmacy [7,15]. In these qualitative studies, many GPs call for better interprofessional communication and a fair balance between them and specialists when sharing prescribing activity [15,16]. Our quantitative study con rmed that cooperation and con dence between physicians is not obvious for a majority of GPs (moderate and low cooperation pro les).
General practitioners would preferentially transfer to nurses, and sometimes pharmacists, advice to patients (lifestyle habits, monitoring of illnesses) and therapeutic education [19,20].
Our results highlight the diversity of GPs' opinions about the role of pharmacists in the management of polypharmacy. Previous studies have demonstrated that pharmacists' involvement in medication management may result in better clinical outcomes and improvement of prescribing patterns, especially for patients with polypharmacy [21][22][23]. Their involvement in patient counselling, therapeutic education or the training of other health professionals would have a positive impact on the patients follow-up, their treatment adherence and their quality of life [22]. In the same time, the frequent limitation of cooperation between GPs and pharmacists to clari cation of prescriptions, drug-related information, or information about patient history reduces its potential bene ts [24][25][26]. Moreover, contacts are occasional and mostly at the pharmacists' initiative. Nevertheless, pharmacists and GPs share the opinion that cooperation is easier when they have a local, long-lasting working relationship [25,27].
Our nding that a majority of GPs were favorable to practice nurses ensuring consultations for patients with chronic diseases is encouraging. A recent Cochrane review showed that nurse practitioners achieved equal or better outcomes for chronic patients than primary care doctors, in terms of quality of care, patient health status, and patient satisfaction [28].
The younger age of GPs with a moderate cooperation pro le, compared with the others, probably re ects the shaping of attitudes and behaviors regarding interprofessional cooperation by years of experience and professional environment: more experienced GPs may have adjusted their role, given what they have learned to expect from other health professionals in their environment.
The absence of association between GPs' pro les and multiprofessional practice organization is surprising. But our study focused especially on the roles of specialists and pharmacists, who are rarely integrated in such organizations in France. Moreover, multiprofessional and group practices themselves have various pro les of organization and interprofessional collaboration [29,30].
We found that GPs with pro les of moderate or selective cooperation had lesser deprescription propensities. At the same time, GPs with low and intensive cooperation pro les had similar behaviors towards deprescription.
Qualitative studies have indicated that some GPs fear con icts with other physicians or pharmacists when they initiate deprescription [15,31]. One hypothesis is that GPs who cooperate intensively discuss and share the responsibility for the prescription management with other professionals and might thus feel themselves supported and con dent when deprescribing. Inversely, GPs with a low cooperation pro le may prefer to take on the responsibility for prescriptions management including when deprescribing. This may be problematic especially when they face therapeutic dilemmas, in which they could bene t from the advice and insight of colleagues and other professionals.

Strengths And Limitations
The size of our sample was substantial. Moreover, we weighted the data according to some GP characteristics to minimize any selection bias. The method we used, AHCA, enabled us to study the statistical proximity of individuals on the basis of the factors under study, without any preconceptions about any relations among these factors. This analysis allows the identi cation of clusters of GPs assessed for several factors.
Among the limitations, GPs' opinions and behaviors related to interprofessional cooperation with and the roles of other healthcare professionals were self-reported. We could not measure real collaboration practices, which would have required a more complex design. Thus, social desirability or conformity biases cannot be ruled out.

Practice and policy implications
Health policies in various countries are promoting interprofessional cooperation [11,13,14] to improve the quality of primary care and patient health, especially for the patient population considered here [32]. Our study highlighted the diversity of GPs' points of view about cooperation with other healthcare professionals and about sharing prescription management for these. Although most GPs were favorable to interprofessional cooperation, they were also largely reluctant to delegate prescription tasks. This re ects both their strong attachment to the symbolic act of prescription and their defensiveness about other professionals in relation to elds of competence and responsibilities [15,24]. Policy makers must be aware of these disparities and di culties.
Better cooperation requires better communication and better understanding of each other's' roles, knowledge, and responsibilities [15,24,27]. Our results indicate that most cooperative GPs have an interest in continuing training. The inclusion of multiprofessional courses in training programs for health students in medicine, pharmacy, nursing, and other disciplines, may lead to a better comprehension of the roles and skills of each profession and of the potential synergies to be gained in interprofessional cooperations [8,33]. The rapid development of multiprofessional medical centers in France and elsewhere is an opportunity to stimulate and facilitate such collaborations.

Conclusion
Disparities exist among GPs regarding the interprofessional cooperation in the management of patients with multimorbidity and polypharmacy. The majority of GPs are inclined to cooperate with other health professionals, but not in the same way, not with the same professionals, and very few go so far as to share prescription management with non-physician professionals. Interprofessional education, whether initial or continuing, may be one way to improve knowledge and understanding of each professional's roles and skills in the management of complex patients.
Abbreviations AHCA agglomerative hierarchical cluster analysis CME continuing medical education GP general practitioner MCA multiple correspondence analysis Declarations Ethics approval and consent to participate:A consent to participate was obtained from each general practitioner at the time of the inclusion in the panel. The National Authority for Statistical Information (Commission Nationale de l'InformationStatistique) approved the panel and its surveys (Paris, June the 3rd 2013, n°82 /H030). This national institution evaluated that the study was in accordance with the rules and regulations regarding the protection of personal data.
Consent for publication:Not applicable.
Availability of data and materials: The datasetsused and analysedduring the currentstudy are availablefrom the correspondingauthor on reasonablerequest. Descriptive analysis of data used in the currentstudy are in an additional le.