Social networks of oncology clinicians as a means for increasing survivorship clinic referral.

Background: Specialized cancer survivorship clinics are recommended for addressing treatment-related health concerns of long-term survivors, but their relative newness in medical oncology necessitates strategies to expand services and clinic referrals. This study used social network analysis to identify personal and/or network factors associated with referral of patients to a survivorship clinic.
Methods: We conducted a cross-sectional social network survey of clinical personnel at a National Cancer Institute-designated comprehensive cancer center. Participants identified colleagues with whom they consult for advice (advice network) and/or discuss patient care (discussion network). Exponential random graph models and logistic regression were used to identify key opinion leaders in the network and factors associated with referral of patients to the center's survivorship clinic.
Results: Here we show that of the respondents (n = 69), 78.0% report being aware of the survivorship clinic, yet only 30.4% had ever referred patients to it. Individuals tend to associate with others in the same occupational role (homophily). In the discussion network, holding an influential network position (betweenness centrality) is associated with patient referral to the survivorship clinic. In the advice network, several social workers are identified as opinion leaders. Conclusions: This study shows that there is strong homophily in both networks, potentially inhibiting information sharing between groups. In designing an inclusive network intervention, persons occupying influential network positions and opinion leaders (e.g., social workers in this case) are well-positioned to promote survivorship clinic referrals.

Introduction

Due to recent advances in diagnosis and therapy, most adults treated for cancer now achieve long-term survival with 84.6% of patients aged 19–39 years, and 72.7% of patients aged 40–64 years, living 5 years or longer[1]. Unfortunately, many survivors develop clinically significant health problems as a result of cancer treatment, resulting in physical symptoms, functional impairment, early mortality, diminished health-related quality of life, emotional distress, barriers to school and work, and financial insecurity[2]. To address these issues, specialized survivorship clinics are recommended as a means for adult cancer survivors to obtain critical services including medical monitoring and management of chronic side effects, psychosocial support, fertility assistance, and written survivorship care plans summarizing prior treatment and recommendations for improving health[3-7].

Cancer survivorship is a relatively new discipline within adult oncology and the availability of specialized clinics for adult cancer survivors is limited. This is in contrast to pediatric oncology, where survivorship clinics are ubiquitous and their role well-established due to historically high long-term survival rates of children with cancer, their high prevalence of late effects, extensive survivorship research, and availability of long-term follow-up guidelines[8-12]. Building on this, the National Comprehensive Cancer Network and the American College of Surgeons have prioritized provision of appropriate adult-focused survivorship care[7,13,14]. Therefore, effective strategies are needed for cultivating survivorship services in adult cancer centers, including optimizing referrals to cancer survivorship clinics, which have been identified as an effective model for delivering survivorship care[6].

However, implementation of new clinical practices and acceptance of new procedures can be a lengthy process[15]. One innovative approach to understanding and potentially impacting the dynamics of referrals to survivorship clinics is through social network analysis. Social network analysis is a scientific discipline that explores communication patterns, diffusion of ideas and innovations, and the adoption of new practices[16]. Social network analysis and constructs have been used in clinical and medical settings since the 1950s to understand the adoption of new practices among physicians[17]. Social network-based interventions have been used to promote changes in guideline compliance, prescribing practices, and implementation of evidence-based medicine among clinicians[18-21]. This methodology is critical to understanding the effects of social influence on the dissemination of new ideas and behaviors in clinical settings and can be used to implement change.

This framework provides insight into communication patterns within networks and identifies central individuals who are best able to influence others to change their behavior[22-24]. These individuals are opinion leaders who interact with many others and/or whom others consult for advice. Professional advice networks typically capture the relationships people seek out when they need guidance from someone whose expertize and knowledge they value on a certain topic[25]. Advice network nominations, often to colleagues of higher rank or prestige, can be used to identify opinion leaders who can act as agents of change in network interventions[16]. In contrast, discussion networks are often to colleagues/peers of similar rank or prestige and capture relationships with individuals to discuss their work, but not necessarily to seek advice or expertize (see Supplementary Table 1 for glossary terminology).

Social network dynamics have been shown to impact health outcomes and utilization of healthcare[26,27]. Social network analysis has been used to characterize provider collaborations in breast cancer care[28] and initiatives for reducing cancer care disparities[29], but not, to our knowledge, in cancer survivorship care. The aim of this study was to identify the opinion leaders in oncology clinician advice and discussion networks regarding referral of patients to a newly-established survivorship clinic at a National Cancer Institute (NCI)-designated comprehensive cancer center. We hypothesized that network members would cluster based on both clinical roles and patient referral patterns. The overall goal was to identify personal and/or network factors that were associated with survivorship clinic referrals and are relevant to development of a network intervention to increase clinic referrals. Results show evidence of role homophily in both the advice and discussion networks, potentially inhibiting the flow of communication between individuals in different clinical roles. Individuals with high betweenness centrality, who occupy bridging positions, were significantly more likely to refer patients. Additionally, social workers emerged as opinion leaders in the advice network and may be influential in promoting clinic referral.

Methods

Study design, setting, participants and procedures

The study comprised a cross-sectional survey of clinicians and clinical support staff who could potentially refer patients to the cancer survivorship clinic at the Norris Comprehensive Cancer Center. Started in 2017, the cancer survivorship clinic provides in-depth, post-treatment assessments for cancer survivors with the goal of improving health outcomes. Referral guidelines specify patients treated with curative intent at less than 50 years of age using cancer therapy associated with long-term toxicity.

Eligible clinicians included treating physicians (medical, surgical, and radiation oncologists), physician assistants, nurse practitioners, oncology clinic nurses, nurse navigators, social workers, and genetic counselors actively practising at the cancer institute. Eligible clinical support staff included schedulers, direct care partners, and clerical referral specialists. All clinicians and staff had regular, direct contact with patients who met cancer survivorship clinic referral guidelines and played a role in the cancer survivorship clinic referral process. Overall, 163 eligible individuals were identified from medical staff lists, department rosters, and managers and invited to participate. Recruitment was purposeful to include a representative range of cancer- and modality-specific treatment teams, departments, and disciplines.

Potential participants were invited by email to complete a confidential online survey through Qualtrics about professional social networks and patient referrals (See Supplementary Note 1). Data collection occurred from June 2018 to August 2018. Participants were given an information sheet and electronically consented to participate in the study and were compensated with a $10 gift card for survey completion. The study was approved by the University of Southern California Institutional Review Board (approval number HS-09-00673).

Measures

Cancer survivorship clinic knowledge and utilization

Participants were asked: (1) if they were aware that the survivorship clinic existed (yes/no/not sure); (2) if they had referred any patients to the survivorship clinic (yes/no/not sure); (3) if they had referred any patients who met clinic guidelines in the past 12 months; and (4) to estimate how many patients they have referred to the clinic in the past 12 months.

Advice network

Participants were asked to name up to seven individuals at the cancer center whom they go to for advice about any aspect of patient care for their cancer patients. Individuals could be from any clinical role, profession, or occupation. Nominations create a connection between two individuals in a network, referred to as a network tie. Advice network nominations were used to create a directed adjacency matrix, where each directed pair of individuals x = 1, if participant i nominated individual j as a person they work with and go to for advice about patient care.

Discussion network

Participants were asked to name up to seven individuals at the cancer center with whom they discussed any aspect of patient care for their cancer patients. Individuals could be from any occupation at the cancer center, not necessarily the same occupation as oneself. The same individuals could be nominated for both the advice and discussion networks. Discussion network nominations were used to create a directed adjacency matrix, where each directed pair of individuals x = 1, if participant i nominated individual j as a person they discuss patient care with.

Network exposure

Network exposure[30] was calculated as the proportion of individuals in one’s network that reported referring patients to the survivorship clinic (e.g., if someone nominated 6 individuals, and 3 of these nominated individuals had referred patients to the clinic, then personal network exposure would be 0.50).

Statistical analysis

Discussion and advice networks were analyzed separately using exponential random graph models (ERGMs)[31,32]. In ERGMs, the dependent variable is the presence or absence of a tie between two people in the network (1 = present, 0 = absent). ERGM estimation and interpretation is similar to logistic regression. However, ERGMs are unique in that they control for dependencies among observations since there are multiple observations for each member of the network. Additionally, individual attribute and network structural effects are estimated in the same model. ERGMs model the probability of a tie in the observed network occurring more or less often than would be expected by chance while controlling for network density (the number of ties in the network) and network structural effects (i.e., reciprocity or mutual connections). Maximum likelihood estimates for the ERGMs were fit separately for advice and discussion networks using Markov Chain Monte Carlo in R (version 4.1.2).

Advice and discussion networks were restricted to people who completed the survey. To explore the relationships among network ties and participant attributes, an effect was added to each model matching for role at the cancer center, awareness of the survivorship clinic (Yes = 1, No = 0) and referral of patients to the survivorship clinic (Yes = 1, No = 0). Structural effects were added for tie reciprocity, and geometrically weighted edgewise shared partnership (Gwesp), where people have indirect ties throughout the network in common, and geometrically weighted outdegree distribution (Gwodegree), a measure for outdegree distribution in the network. Additional combinations of structural terms (Gwesp, Gwodegree, Gwidegree) were tested but model convergence was not reached. In addition to the exponential random graph models, multivariable logistic regression was used to analyze factors associated with referral of patients to the survivorship clinic while controlling for covariates.

Table 1

Participant characteristics (N = 69)a.

	N (%)
Role at cancer center
Scheduler/otherb	22 (31.2)
Physician	20 (29.0)
Clinic nurse	16 (23.2)
Social worker	6 (8.7)
Physician assistant/nurse practitioner	5 (7.3)
Aware of survivorship clinic
Yes	54 (78.3)
No	8 (11.6)
Don’t know	3 (4.4)
Missing	4 (5.8)
Have referred patients to the survivorship clinic
Yes	21 (30.4)
No	39 (56.5)
Don’t know	4 (5.8)
Missing	5 (7.3)

aIncludes only those participants who provided responses for both the advice and discussion networks.

bIncludes genetic counselors, direct care partners, and clerical referral specialists.

Table 2

Exponential random graph models for advice and discussion networks.

	Advice		Discussion
	Estimate (SE)	P-value	Estimate (SE)	P-value
Structural effects
Edges	−4.97 (0.55)	<0.0001	−4.91 (0.56)	<0.0001
Mutuality	0.65 (0.47)	0.16	1.52 (0.36)	<0.0001
Gwesp	1.15 (0.18)	0.0004	1.02 (0.16)	<0.0001
Gwodegree	−1.47 (0.42)	<0.0001	−1.33 (0.41)	0.001
Attribute effects
Physiciansa	−0.02 (0.07)	0.79	0.01 (0.08)	0.93
Nurse practitioner/physician assistant	0.31 (0.15)	0.04	0.37 (0.13)	0.004
Social workers	0.38 (0.10)	0.0003	0.52 (0.14)	0.0002
Scheduler/other	−0.31 (0.15)	0.04	−0.26 (0.13)	0.05
Awareness of survivorship clinic	0.45 (0.45)	0.32	0.59 (0.50)	0.24
Referral to survivorship clinic	0.10 (0.10)	0.31	0.13 (0.10)	0.20
Matched
Node match role at clinic	0.88 (0.18)	<0.0001	0.79 (0.15)	<0.0001
Node match awareness	−0.05 (0.52)	0.93	−0.42 (0.54)	0.43
Node match referrals	0.19 (0.20)	0.33	0.08 (0.17)	0.66

aClinic nurses as reference group.

Table 3

Multivariable logistic regression on patient referrals in discussion network.

	Estimate (SE)	P-value
Network exposure	−0.37 (0.98)	0.71
Betweenness centrality	0.43 (0.19)	0.025
Role at clinica
Physician Assistant/ Nurse Practitioner	−1.00 (1.30)	0.44
Physician	0.007 (0.76)	0.99
Scheduler/Other	−0.02 (0.76)	0.98
Social Worker	−1.44 (1.38)	0.30

*p < 0.05, **p < 0.01, ***p < 0.001.

aNurses as Reference group.