Abstract
Objective
Research to date focused on quantifying team collaboration has relied on identifying shared patients but does not incorporate the major role of communication patterns. The goal of this study was to describe the patterns and volume of communication among care team members involved in treating breast cancer patients.
Materials and Methods
We analyzed 4 years of communications data from the electronic health record between care team members at Vanderbilt University Medical Center (VUMC). Our cohort of patients diagnosed with breast cancer was identified using the VUMC tumor registry. We classified each care team member participating in electronic messaging by their institutional role and classified physicians by specialty. To identify collaborative patterns, we modeled the data as a social network.
Results
Our cohort of 1181 patients was the subject of 322 424 messages sent in 104 210 unique communication threads by 5620 employees. On average, each patient was the subject of 88.2 message threads involving 106.4 employees. Each employee, on average, sent 72.9 messages and was connected to 24.6 collaborators. Nurses and physicians were involved in 98% and 44% of all message threads, respectively.
Discussion and Conclusion
Our results suggest that many providers in our study may experience a high volume of messaging work. By using data routinely generated through interaction with the electronic health record, we can begin to evaluate how to iteratively implement and assess initiatives to improve the efficiency of care coordination and reduce unnecessary messaging work across all care team roles.
Keywords: workflow, social networking, burnout, collaboration, breast cancer
INTRODUCTION
Cancer treatment is complex, requiring multiple treatment modalities delivered by many specialists over an extended period of time. Fragmented care delivery can result in poor communication and a lack of coordination between care providers.1–4 Effective care coordination, supported by health information technology, has been identified as an opportunity to improve treatment outcomes by connecting specialists and care teams to reduce unnecessary complexity.5–7 Care coordination involves the deliberate organization of patient care activities, including effective communication and information transfer, to ensure appropriate health care service delivery.8,9 Previous studies have found that effective care coordination can improve the timeliness of treatment, improve outcomes, and reduce treatment costs.10,11
Previous health care delivery studies have found that care team members, such as patients, caregivers, and clinicians, perform significant invisible work to effectively manage treatments.12,13 Current care coordination efforts commonly rely on a provider-centric model, in which a single physician or small group of physicians key to the patient’s treatment must actively integrate plans across an entire care team.14 Provider-centric care coordination can burden providers with nonclinical administrative and clerical tasks, leading to provider burnout.15–17 Transitioning to a shared care model, in which multiple members of the care team participate in coordination activities, has been identified as 1 solution to maintain effective coordination and improve professional satisfaction.18,19
Studying care coordination and collaboration through the breadth of data collected through routine use of clinical information systems is a critical way to gain insights into clinical practice. The electronic health record (EHR) system is an essential tool for collaboration between clinical personnel.20,21 Functionality such as a shared patient chart and secure clinical messaging encourages communication and teamwork across care team stakeholders to support optimal patient care.20–24 Patient portals and consumer health technologies similarly support collaboration between the patient and their care team outside of the clinical setting.24–28 Using routinely collected data from clinical care to quantify coordination and collaboration between team members can provide feedback to support iterative process improvement.29,30
There are many challenges in understanding the scope of collaborative workload. The importance of care coordination is recognized, but few studies have assessed the scope of the care team involved in coordination activities, such as electronic messaging. Research to date focused on quantifying care team collaboration has relied on identifying shared patients but does not incorporate the major roles of communication patterns.31–34 A previous study by Smith and colleagues found that cancer patients see an average of 32 physicians over the course of their treatment.35 However, we hypothesize that the number of care team members involved in managing, supporting, and delivering breast cancer treatment is much larger. In previous work, we quantified the connectedness among physicians and assessed how connections between physicians changes over time during the treatment of stage I-III breast cancer patients.34,36 We found that medical oncologists are connected to an average of 737 unique appointment providers.36 In this study, we seek to describe communication patterns between members of the cancer care team, including physicians, nurses, clinicians, and administrative staff at a large academic medical center.
MATERIALS AND METHODS
We conducted this study at the Vanderbilt-Ingram Cancer Center at Vanderbilt University Medical Center (VUMC). VUMC is a large academic medical center located in central Tennessee, which provides regional referral care across the southeastern United States. We collected outpatient appointment data and secure clinical messaging data from the EHR on patients who met inclusion criteria for the Vanderbilt University tumor registry: individuals who had been diagnosed or received part of their first course of treatment at VUMC.37 The Vanderbilt University Institutional Review Board approved study procedures.
Study population
We extracted data from the Vanderbilt University tumor registry on patients who were diagnosed with stage I, II, or III breast cancer between January 1, 2014 and December 31, 2016. Tumor registry data included a unique patient identifier, date of diagnosis, and cancer stage. We extracted all outpatient appointment data corresponding to a patient in our cohort from the EHR.38 Appointment data included an appointment date, the provider (eg, physician, physician assistant, or nurse practitioner) associated with each appointment, and the patient identifier. We required that each respective patient have at least 1 outpatient appointment at VUMC in the 6 months following their breast cancer diagnosis to ensure that patients received at least part of their first course of breast cancer treatment at our institution. We also extracted secure clinical communication message logs between January 1, 2014 and November 1, 2017 from the EHR corresponding to each viewed or sent message about a patient in our cohort. Message log data included a unique employee identifier, a unique message thread identifier, a message date and timestamp, and the action performed at each messaging instance (eg, hold for future response, mark as urgent, send). We mapped the employee identifier of each care team member to the respective job description noted in the EHR and classified each job description into 1 of 7 categories (Box 1). We further grouped job classifications into clinical employees and administrative employees. Clinical employees include clinical technicians, nurse practitioners, physician assistants, nurses, and pharmacy staff; administrative employees include administrative and other/nonclinical roles. We kept physicians as their own classification due to their central role in directing breast cancer treatment decisions.
Box 1. Job classifications for VUMC employees involved in secure clinical communications about a patient in our cohort.
Administrative: Individuals who were VUMC employees involved in scheduling or other clinical administrative tasks
Clinical Technicians: Individuals who assisted in patient care, but were not licensed to treat or diagnose a patient
Nurse Practitioner or Physician Assistant: Individuals who were licensed as either a nurse practitioner or physician assistant who were involved in the treatment of a patient in our study cohort
Nurse: Individuals who had received a nursing degree or certification and were involved in the treatment of a patient in our study cohort
Pharmacy Staff: Individuals who were employed as a pharmacist or pharmacy technician
Physician: Licensed individuals who received a doctoral medical degree and were involved in the treatment of a patient in our study cohort
Other/Nonclinical: Individuals, such as nonclinical consultants or volunteers, who did not fit under 1 of the previous classifications.
Network representation
We modeled messaging data as a social network to understand the nature of relationships. We first created patient–employee bipartite networks to understand the relationships between employees and patients and to form a structural foundation for our employee network. Bipartite networks are graphs in which nodes are uniquely members of a distinct set, such that no 2 nodes within the same set are connected.39 In our bipartite network, breast cancer patients in our study population formed 1 set, while employees who were involved in secure clinical messaging about a patient in our population formed a second set. Patient–employee network relationships, or edges, represented the existence of a secure clinical messaging thread in which an employee was involved about a patient in our study population. The size of each patient–employee edge represented the number of unique message threads involving the employee about the respective patient.
To understand relationships between care team members, we created an undirected unipartite projection of the patient–employee network to form an employee–employee network. In the employee–employee network, each node in the graph represented a single employee who sent a message about a patient in our cohort. Edges represented relationships between employees connected by a shared message thread about a unique patient. Both nodes and edges assumed properties, such as size or color, to indicate network characteristics. In our network, edge size uniquely represented 2 separate characteristics: the number of shared patients and the number of mutual message threads. Similarly, node sizes uniquely represented the number of patients about whom the employee communicated and the number of message threads in which the employee was involved. Each employee’s job classification was denoted by the node color.
Data analysis
We created social networks to analyze patient–employee and employee–employee relationships. To assess patient–employee relationships, we calculated descriptive network statistics by 6-month time frames relative to patient diagnosis.36 We assessed employee–employee relationships at 2 levels of granularity. First, we analyzed the employee network across all roles. We also assessed relationships involving providers involved directly in the patient’s breast cancer treatment. We defined medical oncologists, surgical oncologists, radiation oncologists, and plastic surgeons as the cancer providers due to the frequency with which they are involved in treatment. In both employee–employee networks, we calculated descriptive network statistics. All analyses were conducted using the igraph package in R.40,41
We also visualized a clustered network of employees involved in the treatment of patients in our cohort. In the clustered network, we required that each provider pair be involved in at least 50 shared message threads such that we included only employees who were routinely involved in breast cancer care. From this network, we applied a hierarchical clustering algorithm to detect communities of providers whose communication patterns are structurally similar. We chose a hierarchical clustering approach, based on the Girvan-Newman algorithm,42 to community detection as it has been shown to be beneficial to modeling the relationships between individuals within a health care organization.43,44 The Girvan-Newman algorithm organically detects communities from the graph structure, rather than classifying a graph into a user-defined number of clusters. By defining communities based on graph structure, we can gain insight into communication patterns between care team members that would not otherwise be recognizable by other clustering techniques. In the visualized network, we colored edges by community and nodes by employee role to aid in visual analysis. We calculated edge size as the number of shared threads between employee–employee pairs, and node size as the number of threads involving each respective employee. We calculated descriptive statistics for each respective cluster and the network as a whole.
RESULTS
Between January 1, 2014 and December 31, 2016, there were 1557 patients diagnosed with breast cancer at VUMC. Among these patients, 1232 were diagnosed with stage I – III disease. There were 51 patients who did not have an outpatient appointment at our institution within 6 months postdiagnosis, leaving 1181 patients in our study cohort. Each patient averaged 66.2 appointments at our institution during the 24-month study period. Table 1 summarizes network statistics, relative to patient diagnosis date. Patients in our study cohort were the subject of 322 424 messages, sent by 5620 employees in 104 210 unique threads from January 1, 2014 to November 1, 2017. Each patient, on average, was the subject of 88.2 message threads by 106.4 unique employees. The majority of these employees were administrative (53.4) or nonphysician clinicians (42.3). In the first 6 months following diagnosis, there were, an average of 46.2 message threads per patient involving 53.9 employees. Through the patient portal, patients were involved in 27 896 (26.8%) unique message threads.
Table 1.
Patient Messaging Statistics Relative to Diagnosis
| Entire Network | Diagnosis to 6 Months | 6 Months to 1 Year | 1 Year to 18 Months | 18 Months to 2 Years | Over 2 Years | |
|---|---|---|---|---|---|---|
| Number of Patients | 1181 | 1177 | 1088 | 945 | 691 | 529 |
| Number of Appointments | 78 969 | 39 945 | 19 708 | 8131 | 4871 | 5385 |
| Number of Employees | 5620 | 3334 | 2986 | 2751 | 2295 | 2476 |
| Number of Message Threads | 104 210 | 54 372 | 21 335 | 11 757 | 7408 | 9516 |
| Mean Number of Employees per Patient (Range) | 106.4 (2, 363) | 53.9 (2, 126) | 38.4 (2, 92) | 33.4 (2, 107) | 35.1 (2, 109) | 62.0 (2, 206) |
| Physician - Cancer Specialist | 5.6 (1, 10) | 4.5 (1, 8) | 3.7 (1, 8) | 2.7 (1, 6) | 2.6 (1, 7) | 3.1 (1, 9) |
| Physician - Other | 9.4 (1, 31) | 4.4 (1, 15) | 4.0 (1, 17) | 4.2 (1, 14) | 4.2 (1, 15) | 6.1 (1, 16) |
| Clinical Employees | 42.3 (1, 131) | 22.3 (1, 51) | 15.7 (1, 37) | 13.4 (1, 42) | 14.2 (1, 43) | 25.4 (1, 83) |
| Administrative Employees | 53.4 (1, 214) | 25.3 (1, 71) | 17.9 (1, 55) | 16.2 (1, 54) | 16.7 (1, 67) | 30.7 (1, 104) |
| Mean Number of Message Threads per Patient (Range) | 88.2 (1, 821) | 46.2 (1, 291) | 19.6 (1, 134) | 12.4 (1, 136) | 10.7 (1, 134) | 18.0 (1, 318) |
| Physician - Cancer Specialist | 27.1 (1, 210) | 17.5 (1, 91) | 7.1 (1, 47) | 3.4 (1, 28) | 3.3 (1, 55) | 4.7 (1, 95) |
| Physician - Other | 19.9 (1, 226) | 9.0 (1, 65) | 6.2 (1, 59) | 5.6 (1, 57) | 5.7 (1, 65) | 8.8 (1, 93) |
| Clinical Employees | 73.4 (1, 634) | 38.9 (1, 257) | 16.3 (1, 128) | 10.3 (1, 107) | 9.0 (1, 124) | 14.7 (1, 221) |
| Administrative Employees | 66.7 (1, 755) | 33.0 (1, 234) | 15.8 (1, 128) | 10.4 (1, 112) | 9.0 (1, 102) | 15.1 (1, 305) |
Table 2 summarizes network statistics by employee role. There were 1655 administrative staff and 1273 nurses who were involved in communicating about patients in our study, more than any other role. Nurses and physicians were involved in 98% and 44.1% of all messaging threads, respectively. Employees in administrative roles sent more messages (117 473) and were involved in more employee–employee relationships (38 695) than any other specialty. Each administrative employee, on average, communicated about 23 patients, the most of any specialty. Across all roles, each employee was involved in communications about an average of 14.7 unique patients.
Table 2.
Employee Network Statistics by Role
| Administrative | Clinical Technician | Nurse Practitioner or Physician Assistant | Nurse | Pharmacy Staff | Physician (Non-Cancer Provider) | Physician (Cancer Provider) | All Employees | |
|---|---|---|---|---|---|---|---|---|
| Number of Employees | 1655 | 536 | 171 | 1273 | 96 | 1000 | 19 | 5620 |
| Number of Patients | 1181 | 1125 | 1058 | 1173 | 208 | 1014 | 1107 | 1181 |
| Patients per Employee | ||||||||
| Mean (Range) | 23.0 (1, 963) | 12.8 (1, 813) | 18.1 (1, 458) | 18.8 (1, 1030) | 6.6 (1, 83) | 4.6 (1, 216) | 158.7 (6, 398) | 14.7 (1, 1030) |
| Median | 5 | 2 | 3.0 | 4.0 | 2.5 | 2.0 | 116.0 | 3.0 |
| Number of Unique Message Threads (%) | 83 587 (80.2) | 12 658 (12.1) | 13 272 (12.7) | 10 2126 (98.0) | 1210 (1.2) | 21 520 (20.7) | 24 462 (23.5) | 104 210 |
| Message Threads per Employee | ||||||||
| Mean (Range) | 50.5 (1, 3745) | 23.6 (1569) | 77.6 (1, 2516) | 80.2 (1, 10149) | 12.6 (1, 182) | 21.5 (1, 1513) | 1287.5 (16, 4906) | 51.6 (1, 10149) |
| Median | 7.0 | 3.5 | 5.0 | 9.0 | 5.0 | 5.0 | 971.0 | 7.0 |
| Number of Sent Messages (%) | 117 473 (36.4) | 13 130 (4.1) | 17 308 (5.4) | 107 336 (33.3) | 1857 (0.6) | 27 860 (8.6) | 33 805 (10.5) | 322 424 |
| Sent Messages per Employee | ||||||||
| Mean (Range) | 77.6 (1, 6122) | 28.8 (1, 992) | 105.5 (1, 3268) | 94.4 (1, 13063) | 25.4 (1, 368) | 30.0 (1, 1792) | 17 79.2 (14, 7585) | 72.9 (1, 13063) |
| Median | 12.0 | 6.0 | 7.5 | 11.0 | 11.0 | 7.0 | 1178.0 | 9.0 |
| Number of Employee Relationships (%) | 38 695 (33.1) | 10 317 (8.8) | 4598 (3.9) | 35 071 (30.0) | 1426 (1.2) | 15 128 (12.9) | 4727 (4.0) | 117 026 |
| Relationships per Employee | ||||||||
| Mean (Range) | 24.4 (1, 411) | 17.9 (1, 279) | 34.2 (1, 660) | 28.0 (1, 884) | 15.2 (1, 119) | 15.6 (1, 255) | 248.8 (23, 668) | 24.6 (1, 884) |
| Median | 10.0 | 6.0 | 9.0 | 12.0 | 7.0 | 8.0 | 246.0 | 9.0 |
There were 24 providers in our network who were related directly to breast cancer treatments. These providers were involved in 30 242 unique message threads, which accounted for 65.8% of all message threads sent by physicians. Table 3 presents the cancer provider network statistics. Each medical oncologist, on average, communicated with 428.6 other providers in 2230 different message threads, more than providers of any other specialty. Among all specialties, radiation oncologists had on average the fewest number of connections per provider (108) and the fewest number of message threads per provider (344). Across all 4 specialties, providers had more connections with administrative employees than any other employee group. With the exception of plastic surgery, providers shared the most patients and messaging threads with clinical employees. Among medical oncologists, 58.2% of the shared threads involved clinical employees.
Table 3.
Cancer Provider Network Statistics. We include physicians, nurse practitioners, and physician assistants in each respective specialty since individuals in all 3 roles serve in a provider capacity
| Medical Oncology | Surgical Oncology | Radiation Oncology | Plastic Surgery | |
|---|---|---|---|---|
| Number of Providers | 8 | 9 | 3 | 4 |
| Number of Messaging Threads | 17 078 | 10 786 | 1024 | 3025 |
| Number of Patients | 1006 | 915 | 405 | 449 |
| Number of Messaging Threads per Provider | ||||
| Mean (Range) | 2233.0 (388, 4905) | 1417.8 (244, 2514) | 344 (16, 971) | 761.5 (32, 1754) |
| Median | 2214 | 1013 | 45 | 630 |
| Number of Patients per Provider | ||||
| Mean (Range) | 257.9 (43, 449) | 214.3 (45, 424) | 139.7 (8, 382) | 124.8 (6, 309) |
| Median | 263.5 | 223 | 29 | 92 |
| Number of Connections per Provider | ||||
| Mean (Range) | 428.6 (186, 689) | 271.3 (125, 373) | 108 (24, 264) | 170.5 (26, 275) |
| Median | 452.5 | 296 | 36 | 190.5 |
| Total Number of Connections | 3381 | 2394 | 321 | 676 |
| Cancer Specialists (%) | 143 (4.2) | 145 (6.1) | 26 (8.1) | 50 (7.4) |
| Other Physicians (%) | 549 (16.2) | 279 (11.7) | 57 (17.8) | 47 (7.0) |
| Clinical Employees (%) | 1314 (38.9) | 912 (38.1) | 115 (35.8) | 244 (36.1) |
| Administrative Employees (%) | 1375 (40.7) | 1059 (44.2) | 123 (38.3) | 335 (49.6) |
| Total Number Shared Patients | 21 112 | 16 349 | 1699 | 4176 |
| Cancer Specialists (%) | 1567 (7.4) | 2249 (13.8) | 260 (15.3) | 275 (6.6) |
| Other Physicians (%) | 979 (4.6) | 529 (3.2) | 148 (8.7) | 114 (2.7) |
| Clinical Employees (%) | 10 161 (48.1) | 6890 (42.1) | 816 (48.0) | 1451 (34.7) |
| Administrative Employees (%) | 8405 (39.8) | 6681 (40.9) | 475 (28.0) | 2336 (55.9) |
| Total Number of Shared Threads | 35 346 | 24 445 | 2005 | 6464 |
| Cancer Specialists (%) | 2287 (6.5) | 3669 (15.0) | 308 (15.4) | 352 (5.4) |
| Other Physicians (%) | 1324 (3.7) | 1110 (4.5) | 184 (9.2) | 149 (2.3) |
| Clinical Employees (%) | 20 556 (58.2) | 11 426 (46.7) | 983 (49.0) | 2394 (37.0) |
| Administrative Employees (%) | 11 179 (31.6) | 8240 (33.7) | 530 (26.4) | 3569 (55.2) |
Figure 1 presents the clustered network of individual pairs exchanging at least 50 messages. Across all clusters, there were 90 unique employees who were each connected to an average of 3.9 other care team members. Twenty-three of the employees were physicians, 31 were nonphysician clinical employees, and 34 were administrative staff. Individual cluster statistics are presented in Table 4. The cluster involving 7 medical oncologists had 29 unique employees and 86 intracluster edges, the most of any cluster. The cluster involving 7 surgical oncologists had the most edges with other clusters (63). Network density was smallest in the cluster containing 1 medical oncologist and 20 supporting staff. The 90 employees in the clustered network were involved in 71 609 (68.7%) of the unique message threads for this patient population.
Figure 1.
Clustered Network. Each cluster is visualized by a unique edge color. Grey edges represent connecting ties between providers involved in different clusters
Table 4.
Network Cluster Statistics
| Cluster Color | Cluster Description | Number of Employees | Number of Edges to Other Clusters | Number of Intra-Cluster Edges | Mean Intra-Cluster Node Degree | Cluster Density |
|---|---|---|---|---|---|---|
| Aqua | Single Medical Oncologist and Supporting Staff | 21 | 34 | 20 | 1.9 | 9.5% |
| Orange | Seven Medical Oncologists and Supporting Staff | 29 | 59 | 86 | 5.9 | 21.2% |
| Blue | Seven Surgical Oncologists and Supporting Staff | 19 | 63 | 52 | 5.5 | 30.4% |
| Pink | Two Surgical Oncologists, Radiation Oncologists, and Supporting Staff | 8 | 15 | 7 | 1.75 | 25.0% |
| Green | Two Plastic Surgeons and Supporting Administrators | 6 | 5 | 5 | 1.7 | 33.3% |
| Yellow | One Plastic Surgeon and Supporting Staff | 7 | 6 | 6 | 1.7 | 28.6% |
DISCUSSION
We conducted a descriptive study to quantify the scope of care team communication during the treatment of breast cancer patients from EHR-based secure messaging data. We performed a social network analysis of routinely collected secure clinical messaging data to investigate care team connectedness across an entire institution. There are few previous studies that have investigated communication patterns of clinical care teams. Those studies have primarily applied qualitative methods, such as observations and interviews or surveys, which are difficult to scale across an entire institution. In 1 study, researchers surveyed medical oncologists to assess their communication with primary care providers.45 Other studies have applied interviews to investigate the importance of communication across care stakeholders, including patients and caregivers, nurses, and multidisciplinary care team members.4,5,11,46,47 Few studies have applied social network analysis to quantify care team connectedness. However, those studies are primarily limited to individuals who bill for patient care. In our previous work, we investigated connectedness between physicians using data from an institutional tumor registry and appointment scheduling data.34,36 However, these data did not include nonbilling members of the care team. Other studies have applied social network analysis to payor databases, such as SEER Medicare, to investigate relationships among providers treating shared patients at multiple institutions.2,31,48
This is 1 of the first studies to quantify the scope of communication of the entire clinical care team across an institution, including individuals who do not directly provide patient care. Electronic health record data sources allowed us to evaluate a broad range of employees, extending the breadth of single payor data at a single institution. Previous work has found that the EHR supports a significant amount of care coordination between care team members at a single institution.49 Our analysis was enabled by the use of secure clinical messaging data, which allowed us to identify all individuals across any institutional role who sent a message about a patient in our cohort. Electronic messaging data from the EHR are unique in that they detail communications between both individuals who are involved in patient care and individuals who support patient care. In 1 previous study, researchers found that patients interact with an average of 32 providers over the course of their cancer treatment.35 As we hypothesized, our results suggest the care team may be more expansive with nearly half of the care team in nonclinical roles.
Our results indicate that breast cancer patients, on average, are the subject of communications by 106 unique employees over the course of 66 appointments. Over half of these employees are in administrative roles, while only 15 are physicians. These results suggest that members of the care team perform a high volume of messaging work to coordinate treatments. Not surprisingly, the first 6 months following diagnosis involved the largest number of employees in communications. This period of intensive treatment can be burdensome to the patient, particularly when they are required to play a role in logistical coordination.50–52 Previous work has found that patients who experience a high treatment burden often suffer from poor adherence to treatments and a diminished quality of life.53,54
Our scalable approach is not without limitations. Our data detail communications between employees at a single academic medical center. Previous work has found that breast cancer patients often receive care from providers across multiple institutions.52,55 However, our results indicate that patients in our study have an average of over 66 appointments in the study period, suggesting that they receive a significant portion of their breast cancer treatment at our institution. In our analysis, we focused on breast cancer patients who were recorded in our institutional tumor registry, which represents a subset of the breast cancer population at VUMC. We chose this patient population such that we could categorize communication patterns specifically for breast cancer-related treatment. However, in making this decision, we excluded many patients who interacted with breast cancer care team members but did have stage I-III disease or did not meet inclusion criteria for the tumor registry. In 1 recent study, Tai-Seale and colleagues found that physicians at their institution were involved in an average of 129 provider–provider messages per week.56 Our future work will aim to quantify the full scope of messaging involvement for care team members by role and specialty. We also do not account for other types of communication such as tumor boards, phone calls, in person conversations, and email messages. At VUMC, EHR-based secure clinical messaging is the preferred form of electronic communication between care team members, but this is not the case for all institutions. We speculate that the longstanding role of messaging as a preferred means of communication between ambulatory clinics at VUMC,57,58 our institutional reliance on secure messaging as a means of documenting discussions between care team members, and the complexity of cancer care are key contributors to the substantial volume of messaging identified in our analysis. In future work we could also analyze EHR artifacts, such as EHR access transaction logs, to understand care team members who may communicate passively by viewing each other’s clinical documentation. Finally, future work will also apply qualitative methods to understand contextual factors contributing to the communication patterns identified in this study. One such study could use results from our messaging analysis to direct a sampling plan to combine clinic observations and semistructured interviews to query perceptions of burnout among care team members. These results could be compared to the messaging statistics and communication patterns to develop a deeper understanding of how message work relates to professional burnout.
We found that nurses were involved in 98% of all message threads, which suggests their significant role in care coordination for breast cancer patients. Similarly, each nurse had an average of 80 relationships with other employees. We hypothesize that clinicians who message frequently are particularly susceptible to burnout from messaging work. There have been many studies that have assessed physician burnout due to administrative or clerical tasks.15,17 However, there has been substantially fewer studies to evaluate the messaging work on nurses and the respective effects on burnout and job satisfaction, despite literature suggesting that burnout is equally prevalent among clinical roles.59 This study serves as a formative step in such an evaluation. We also found that each employee has an average of 24.6 different collaborators. This number is substantially larger for cancer providers, with 428.6 and 271.3 collaborators for medical oncologists and surgical oncologists, respectively. We hypothesize that such extensive messaging by breast cancer care team members is an artifact of the complex, primarily outpatient, care required for breast cancer treatments.
Recognizing employees who commonly communicate about a single patient population allows us to infer opportunities to improve clinic structure such that we can enhance long-term chronic disease treatment and follow-up. In our cluster analysis, we identified 6 clusters of distinct subteams related by specialty who combine to create the larger breast cancer care team. These teams consisted of 90 employees who were routinely involved in communicating about patients in our study cohort. These providers were involved in nearly 70% of the total message threads. The employees across clusters are commonly located in different clinics, but rely on asynchronous communication to coordinate care. We hypothesize that colocation of multiple specialists and support staff within a single clinic could help to improve communication as has been suggested in previous work.60
CONCLUSION
Measuring the scope of communication and connectivity among care team members affords the opportunity to systematically evaluate teams that perform a high volume of messaging work. In this study, we used secure communication logs from the EHR to conduct a descriptive social network analysis of employees communicating about a patient during their breast cancer treatment. Our ability to quantify care team connectivity across an institution is a formative step towards understanding how team structures relate to job satisfaction, burnout, and treatment outcomes. Our results suggest that many care team members, including physicians and nurses, in our study are required to perform a substantial amount of messaging work to coordinate and deliver breast cancer treatments. By better understanding communication and collaboration patterns among care team members, we can begin to evaluate and implement initiatives to reduce provider workload.
FUNDING
Bryan Steitz was supported by the T15LM007450 training grant from the United States National Library of Medicine.
AUTHOR CONTRIBUTIONS
All authors conceived the study design. BS extracted data, performed analysis, and drafted the initial version of the manuscript. KU and ML provided study guidance, critically reviewed each draft, and provided feedback throughout the study.
Conflict of interest statement
None declared.
REFERENCES
- 1. Coleman EA. Falling through the cracks: challenges and opportunities for improving transitional care for persons with continuous complex care needs. J Am Geriatr Soc 2003; 51 (4): 549–55. [DOI] [PubMed] [Google Scholar]
- 2. Hussain T, Chang HY, Veenstra CM, et al. Collaboration between surgeons and medical oncologists and outcomes for patients with stage III colon cancer. J Oncol Pract 2015; 11 (3): e388–97. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Bayliss EA, Ellis JL, Shoup JA, et al. Effect of continuity of care on hospital utilization for seniors with multiple medical conditions in an integrated health care system. Ann Fam Med 2015; 13 (2): 123–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Lancaster G, Kolakowsky-Hayner S, Kovacich J, et al. Interdisciplinary communication and collaboration among physicians, nurses, and unlicensed assistive personnel. J Nurs Scholarsh 2015; 47 (3): 275–84. [DOI] [PubMed] [Google Scholar]
- 5. Easley J, Miedema B, Carroll JC, et al. Coordination of cancer care between family physicians and cancer specialists: importance of communication. Can Fam Physician 2016; 62: e608–15. [PMC free article] [PubMed] [Google Scholar]
- 6. Institute of Medicine (US) Committee on Quality of Health Care in America. Crossing the Quality Chasm: A New Health System for the 21st Century. Washington, DC: Institute of Medicine (US) Committee on Quality of Health Care in America; 2001. [Google Scholar]
- 7. Ruth JL, Geskey JM, Shaffer ML, et al. Evaluating communication between pediatric primary care physicians and hospitalists. Clin Pediatr (Phila) 2011; 50 (10): 923–8. [DOI] [PubMed] [Google Scholar]
- 8.Agency for Healthcare Research and Quality (AHRQ). Closing the Quality Gap: A Critical Analysis of Quality Improvement Strategies, Volume 7–Care Coordination. Rockville, MD: AHRQ; 2007. [PubMed] [Google Scholar]
- 9. Schultz EM, Pineda N, Lonhart J, et al. A systematic review of the care coordination measurement landscape. BMC Health Serv Res 2013; 13 (1): 443–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Peikes D, Chen A, Schore J, et al. Effects of care coordination on hospitalization, quality of care, and health care expenditures among Medicare beneficiaries. JAMA 2009; 301 (6): 603–16. [DOI] [PubMed] [Google Scholar]
- 11. Gorin SS, Haggstrom D, Han PKJ, et al. Cancer care coordination: a systematic review and meta-analysis of over 30 years of empirical studies. Ann Behav Med 2017; 51: 1–15. [DOI] [PubMed] [Google Scholar]
- 12. Ancker JS, Witteman HO, Hafeez B, et al. The invisible work of personal health information management among people with multiple chronic conditions: qualitative interview study among patients and providers. J Med Internet Res 2015; 17 (6): e137–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Bodenheimer T. Coordinating care: a perilous journey through the health care system. N Engl J Med 2008; 358 (10): 1064–71. [DOI] [PubMed] [Google Scholar]
- 14. Press MJ. Instant replay: a quarterback's view of care coordination. N Engl J Med 2014; 371 (6): 489–91. [DOI] [PubMed] [Google Scholar]
- 15. Shanafelt TD, Gradishar WJ, Kosty M, et al. Burnout and career satisfaction among US oncologists. J Clin Oncol 2014; 32 (7): 678–86. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Shanafelt TD, Boone S, Tan L, et al. Burnout and satisfaction with work-life balance among US physicians relative to the general US population. Arch Intern Med 2012; 172 (18): 1377–9. [DOI] [PubMed] [Google Scholar]
- 17. Shanafelt TD, Dyrbye LN, Sinsky C, et al. Relationship between clerical burden and characteristics of the electronic environment with physician burnout and professional satisfaction. Mayo Clin Proc 2016; 91 (7): 836–48. [DOI] [PubMed] [Google Scholar]
- 18. Sinsky CA, Willard-Grace R, Schutzbank AM, et al. In search of joy in practice: a report of 23 high-functioning primary care practices. Ann Fam Med 2013; 11 (3): 272–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. D’afflitti J, Lee K, Jacobs M, et al. Improving provider experience and increasing patient access through nurse practitioner–physician primary care teams. J Ambul Care Manag 2018; 41 (4): 308–13. [DOI] [PubMed] [Google Scholar]
- 20. Chase DA, Ash JS, Cohen DJ, et al. The EHR's roles in collaboration between providers: A qualitative study. AMIA Annual Symposium Proceedings 2014; 2014: 1718–27. [PMC free article] [PubMed] [Google Scholar]
- 21. Sharma N, O'Hare K, O'Connor KG, et al. Care coordination and comprehensive electronic health records are associated with increased transition planning activities. Acad Pediatr 2018; 18 (1): 111–8. [DOI] [PubMed] [Google Scholar]
- 22. Berg M, Bowker G.. The multiple bodies of the medical record: toward a sociology of an artifact. Sociol Q 1997; 38 (3): 513–37. [Google Scholar]
- 23. Pratt W, Reddy MC, McDonald DW, et al. Incorporating ideas from computer-supported cooperative work. J Biomed Inform 2004; 37 (2): 128–37. [DOI] [PubMed] [Google Scholar]
- 24. Bates DW, Bitton A.. The future of health information technology in the patient-centered medical home. Health Aff 2010; 29 (4): 614–21. doi: 10.1377/hlthaff.2010.0007 [DOI] [PubMed] [Google Scholar]
- 25. Darkins A, Ryan P, Kobb R, et al. Care coordination/home telehealth: the systematic implementation of health informatics, home telehealth, and disease management to support the care of veteran patients with chronic conditions. Telemed e-Health 2008; 14 (10): 1118–26. [DOI] [PubMed] [Google Scholar]
- 26. Flatley Brennan P, Valdez R, Alexander G, et al. Patient-centered care, collaboration, communication, and coordination: a report from AMIA's 2013 Policy Meeting. J Am Med Assoc 2015; 22 (e1): e2–e6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Dixon BE, Embi PJ, Haggstrom DA.. Information technologies that facilitate care coordination: provider and patient perspectives. Transl Behav Med 2018; 8 (3): 522–5. [DOI] [PubMed] [Google Scholar]
- 28. Rosenbloom ST, Steitz BD, Warner JL.. Window of opportunity: patient portals and cancer. J Oncol Pract 2018; 14 (11): 639–41. [DOI] [PubMed] [Google Scholar]
- 29. Deeny SR, Steventon A.. Making sense of the shadows: priorities for creating a learning healthcare system based on routinely collected data. BMJ Qual Saf 2015; 24 (8): 505–15. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Budrionis A, Bellika JG.. The learning healthcare system: where are we now? A systematic review. J Biomed Inform 2016; 64: 87–92. [DOI] [PubMed] [Google Scholar]
- 31. Pollack CE, Weissman GE, Lemke KW, et al. Patient sharing among physicians and costs of care: a network analytic approach to care coordination using claims data. J Gen Intern Med 2013; 28 (3): 459–65. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Barnett ML, Christakis NA, O’Malley J, et al. Physician patient-sharing networks and the cost and intensity of care in US hospitals. Med Care 2012; 50 (2): 152–60. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Landon BE, Keating NL, Barnett ML, et al. Variation in patient-sharing networks of physicians across the United States. JAMA 2012; 308 (3): 1–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Steitz BD, Levy MA.. A social network analysis of cancer provider collaboration. AMIA Annual Symposium Proceedings 2016; 2016: 1987–96. [PMC free article] [PubMed] [Google Scholar]
- 35. Smith SDM, Nicol KM, Devereux J, et al. Encounters with doctors: quantity and quality. Palliat Med 1999; 13 (3): 217–23. [DOI] [PubMed] [Google Scholar]
- 36. Steitz BD, Levy MA.. Temporal and a temporal provider network analysis in a breast cancer cohort from an Academic Medical Center (USA). Informatics 2018; 5 (3): 34. [Google Scholar]
- 37.Tennessee Cancer Registry. https://www.tn.govhealthhealth-program-areastcr.html. https://www.tn.gov/health/health-program-areas/tcr.html Accessed October 2018.
- 38. Danciu I, Cowan JD, Basford M, et al. Secondary use of clinical data: the Vanderbilt approach. J Biomed Inform 2014; 52: 28–35. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Newman M. Networks: An Introduction. New York: Oxford University Press; 2010. [Google Scholar]
- 40. Csardi G, Nepusz T. The igraph software package for complex network research. InterJ Complex Syst2006; 1695: 1–9 http://igraph.org.
- 41.RCT. R: A language and environment for statistical computing. https://www.R-project.org/. (accessed March 15, 2019).
- 42. Girvan M, Newman M.. Community structure in social and biological networks. Proc Natl Acad Sci 2002; 99 (12): 7821–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43. Hripcsak G, Vawdrey DK, Fred MR, et al. Use of electronic clinical documentation: time spent and team interactions. J Am Med Assoc 2011; 18 (2): 112–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44. Lurie SJ, Fogg TT, Dozier AM.. Social network analysis as a method of assessing institutional culture: three case studies. Acad Med 2009; 84 (8): 1029–35. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45. Klabunde CN, Haggstrom D, Kahn KL, et al. Oncologists’ perspectives on post-cancer treatment communication and care coordination with primary care physicians. Eur J Cancer Care 2017; 26 (4): e12628–18. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46. Saini KS, Taylor C, Ramirez AJ, et al. Role of the multidisciplinary team in breast cancer management: results from a large international survey involving 39 countries. Ann Oncol 2012; 23 (4): 853–9. [DOI] [PubMed] [Google Scholar]
- 47. Haynes K, Ugalde A, Whiffen R, et al. Health professionals involved in cancer care coordination: nature of the role and scope of practice. Collegian 2018; 25 (4): 395–400. 10.1016/j.colegn.2017.10.006 [Google Scholar]
- 48. Brunson JC, Laubenbacher RC.. Applications of network analysis to routinely collected health care data: a systematic review. J Am Med Assoc 2018; 25 (2): 210–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49. Aarts J, Ash J, Berg M.. Extending the understanding of computerized physician order entry: Implications for professional collaboration, workflow and quality of care. Int J Med Inform 2007; 76: S4–13. [DOI] [PubMed] [Google Scholar]
- 50. Kamal KM. A systematic review of the effect of cancer treatment on work productivity of patients and caregivers. J Manag Care Spec Pharm 2017; 23 (2): 1–27. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 51.Cheng AC, Levy MA. Data Driven Approach to Burden of Treatment Measurement: A Study of Patients with Breast Cancer. AMIA Annu Symp Proc 2016; 2016: 1756–63. [PMC free article] [PubMed]
- 52. Cheng AC, Levy MA.. Determining burden of commuting for treatment using online mapping services-a study of breast cancer patients. AMIA Annual Symposium Proceedings 2017; 2017: 555–64. [PMC free article] [PubMed] [Google Scholar]
- 53. Tran V-T, Montori VM, Eton DT, et al. Development and description of measurement properties of an instrument to assess treatment burden among patients with multiple chronic conditions. BMC Med 2012; 10 (1): 68–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 54. Tran V-T, Barnes C, Montori VM, et al. Taxonomy of the burden of treatment: a multi-country web-based qualitative study of patients with chronic conditions. BMC Med 2015; 13 (1): 115. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 55.Bridewell W, Das AK. Social network analysis of physician interactions: the effect of institutional boundaries on breast cancer care. AMIA Annu Symp Proc 2011; 2011: 152–60. [PMC free article] [PubMed]
- 56. Tai-Seale M, Dillon EC, Yang Y, et al. Physicians’ well-being linked to in-basket messages generated by algorithms in electronic health records. Health Aff 2019; 38 (7): 1073–8. [DOI] [PubMed] [Google Scholar]
- 57. Giuse DA. Supporting communication in an integrated patient record system. AMIA Annual Symposium Proceedings 2003; 2003: 1065. [PMC free article] [PubMed] [Google Scholar]
- 58. Denny JC, Giuse DA, Jirjis JN.. The Vanderbilt experience with electronic health records. Semin Colon Rectal Surg 2005; 16 (2): 59–68. [Google Scholar]
- 59. Dyrbye LN, Shanafelt TD, Sinsky CA, et al. Burnout among health care professionals: a call to explore and address this under recognized threat to safe, high-quality care. NAM Perspect 2017; 7 (7): 1–11. [Google Scholar]
- 60. Reid RJ, Coleman K, Johnson EA, et al. The group health medical home at year two: cost savings, higher patient satisfaction, and less burnout for providers. Health Aff 2010; 29 (5): 835–43. [DOI] [PubMed] [Google Scholar]