All tasks and interaction taking place between agents in executing these tasks are described in this section. Further, the entire hospital network, which is built up by these interactions, is presented and integration and differentiation are described. We start this section with some key figures on surgeries in Slingeland Hospital.
- Output of the social network
In 2017, 10,157 surgeries were performed in Slingeland Hospital. The number of surgeries varies from a minimum of 4 to a maximum of 246 surgeries a week. Of all surgeries, 83% are planned beforehand, i.e., they are not emergency procedures. Different types of surgeries are performed, which are registered according to 394 treatment codes in the HIS. These treatment codes are divided among nine medical disciplines: general surgery, orthopedics, Ear Nose Throat (ENT) surgery, eye surgery, urology, gynecology, plastic surgery, dental surgery and neurosurgery. Of all 394 treatment codes, on average 66% are performed once a month or less and 9% are performed on a weekly basis. More than half of the treatment codes are executed by only one or two specific surgeons. For example, 103 treatment codes are performed by one specific, but not the same, surgeon. For 42% of all surgeries, there was a unique one time combination of treatment code, surgeon and anesthesiologist. This and the fact that in 2017 a total of 2,881 unique combinations of medical instrument sets were used, suggest that human and material resources are not fit for a large variety of surgeries, but are mostly suitable for specific surgeries.
3.2 Agents and tasks performed for surgery
The main task of the logistical system is to get the right patient, surgeon, anesthesiologist, nurses, materials and infrastructure together at the right time and in the right place. There are 23 tasks that are executed in order to achieve this, as presented in Table 2. Figure 2 shows the relation between 22 tasks, mostly based on the chronological order in which these are executed. In addition, the arrow between two tasks means that the output of a task is input to the task to which it is connected. Task 23, managing the OTC, is not specifically time dependent, nor is there specific output of this task and therefore it is not mentioned in Figure 2. Tasks 1 to 5 are at the tactical level because these concern master scheduling in the medium-term [41]. The other tasks are operational because they are related to short-term allocation of resources and execution. Long-term strategic tasks such as demand forecasting were found, but these do not relate directly to the tasks shown in Figure 2. Overall, two main groups of tasks are visible in Figure 2: tactical and operational planning 6 months ahead until the day before surgery and the execution of the surgery process.
First the OR master schedule is made for a three-month period (task 1), two quarters ahead; the OR master schedule for Q2 of any year is made in Q4 of the previous year. In the OR master schedule, time slots for all ORs are allocated to the nine medical disciplines that operate in the OTC. The clinical bed plan (task 2), equipment maintenance planning (task 5) and staff schedules (tasks 3 and 4) are all derived from the OR master schedule. Around two to twelve weeks before surgery, patients are planned into the OR program (task 6) and preparations start: patients are screened by an anesthesiologist (task 8), materials are ordered (task 7) and patients are seen by other physicians or take radiology or laboratory tests (task 9). In the days before surgery further preparations are made: the OR day program is planned in more detail (task 11), staff is allocated to specific surgeries (task 10) and materials are picked (task 12). On the day of surgery the patient is prepared and held on the ward (task 14) before the actual surgery takes place (task 17) and is afterwards taken care of in the recovery area (task 20) and ward (task 21). In some cases a radiology image is made during or after surgery (task 16). After surgery the OR is cleaned (task 18) and if necessary the medical instruments are immediately cleaned for reuse (task 19). Patients can also be admitted for an emergency surgery (task 13), in which case all tasks are executed within a short period of time. All tasks have been specified in more detail in Additional File 2.
Tasks are related to patient, staff and material flows. Tasks 5, 7, 12, 18 and 19 are related to materials and tasks 1, 3, 4 and 10 are about staff flows. Tasks 2, 8, 9, 13 and 16 are related to patient flow. The other tasks are related to more than one flow; for example, preparing a patient on the ward before surgery involves both the patient and medication.
For each task a number of agent types is involved, as presented in Table 2. The OTC day coordinator participates in 11 of the tasks and has the highest involvement in multiple tasks. The OTC capacity planner, anesthesiologists, surgeons and the OTC nurses all participate in eight different tasks. The other agents participate in fewer tasks, with a minimum of one. The task with the most different agent types involved is managing the OTC day program; ten different agent types play a role in this.
With regard to the flows, most agents are involved in tasks related to patients, staff and materials. The CSD staff members, equipment maintenance staff, OTC cleaning staff and the OTC logistical staff are the only agents who deal with just one flow type, which is materials.
Table 2 shows that a number of tasks have an overlap in types of participating agents. This is particularly relevant when tasks are related. For example, tasks 2, 3, 4 and 5 are all related to task 1, and there is overlap in agent participation for the OTC capacity planner for tasks 1 and 2. The outpatient secretary, the anesthesiologist and the surgeon are all involved in both tasks 1 and 3. There are no overlapping agents for related tasks 1 and 4, 6 and 8, 6 and 14, 13 and 14 and 13 and 15. For tasks 4, 8 and 14 relevant information resulting from tasks 1 and 6 are communicated through the HIS, which then is the only information source for agents executing these tasks. For all other related pairs of tasks there are agents who participate in both tasks.
Table 2 Tasks and agent types involved in these tasks
MSC = Medical Specialty Company; CSD = Central Sterilization Department; MIS = Medical Instrumental Services; OTC = Operating Theatre Complex; ER = Emergency Room department; Preop dpt. = Preoperative Screening department; (n) = number of departments
|
Tasks
|
MSC
|
OTC
|
MSC
|
Cluster OTC and Services
|
CSD
|
ER
|
ER
|
MIS
|
OTC
|
OTC
|
OTC
|
OTC
|
OTC
|
OTC
|
OTC
|
OTC
|
Outpatient dpt (5)
|
Pharmacy dpt
|
Preop dpt
|
Preop dpt
|
Radiology dpt
|
OTC
|
MSB
|
Nrusing ward (7)
|
Nursing ward A2
|
Nursing ward (7)
|
Task description
|
|
Anesthesiologist
|
Nurse anesthetist
|
Assistant surgeon
|
Cluster manager
|
CSD staff member
|
ER physician
|
ER nurse
|
Equipment maintenance staff
|
Holding nurse
|
OTC capacity planner
|
OTC cleaning staff
|
OTC day coordinator
|
OTC Logistical staff
|
OR nurse
|
OTC secretary
|
OTC team leader
|
Outpatient secretary
|
Pharmacy assistant
|
Preoperative nurse
|
Preoperative secretary
|
Radiology staff
|
Recovery nurse
|
Surgeon
|
Ward nurse
|
Clinical bed plan boss
|
Ward team leader
|
|
1
|
Make OR master schedule
|
x
|
|
|
x
|
|
|
|
|
|
x
|
|
|
|
|
|
|
x
|
|
|
|
|
|
x
|
|
|
|
Make the OR master schedule in which operating time for each medical discipline is allocated to the operating rooms for even and uneven weeks
|
|
2
|
Make clinical bed plan
|
|
|
|
|
|
|
|
|
|
x
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
x
|
x
|
Make the clinical bed plan in which beds are allocated to medical disciplines per nursing ward for even and uneven weeks.
|
|
3
|
Schedule surgeons and anesthesiologists
|
x
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
x
|
|
|
|
|
|
x
|
|
|
|
Determine the working hours for every surgeon and anesthesiologist for the upcoming three to six months, including where they are working, i.e., in the outpatient department and the OTC.
|
|
4
|
Schedule OTC nurses
|
|
X
|
|
|
|
|
|
|
x
|
|
|
x
|
|
X
|
|
x
|
|
|
|
|
|
x
|
|
|
|
|
Determine the working hours for every nurse anesthetist, OR, holding and recovery nurse for the upcoming three to six months.
|
|
5
|
Plan equipment maintenance
|
|
|
|
|
|
|
|
x
|
|
x
|
|
x
|
|
|
|
x
|
|
|
|
|
|
|
|
|
|
|
Determine which OTC equipment will be maintained on what day and time and how long the equipment will be unavailable for use.
|
|
6
|
Plan surgery
|
|
|
|
|
|
|
|
|
|
x
|
|
|
|
|
|
|
x
|
|
|
|
|
|
x
|
|
x
|
|
Determine the time and date that the patient will be operated on and register this in the OR master schedule.
|
|
7
|
Order materials
|
|
|
|
|
|
|
|
|
|
|
|
x
|
|
X
|
|
|
x
|
|
|
|
|
|
x
|
|
|
|
Request specific materials for one specific surgery and order these materials at external supplier(s)
|
|
8
|
Preoperative screening
|
x
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
x
|
x
|
x
|
|
|
|
|
|
|
Determine what type of anesthetic technique fits the patient, what potential risks should be considered during the patient’s surgery, and what is the best preparation for the surgery of this patient.
|
|
9
|
Make appointment
|
x
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
x
|
|
|
x
|
|
|
|
|
|
|
Request visits to physicians, laboratory or radiology tests for preparation of the patient for the surgery.
|
|
10
|
Plan OTC nurses
|
|
X
|
|
|
|
|
|
|
x
|
|
|
x
|
|
X
|
|
|
|
|
|
|
|
x
|
|
|
|
|
Determine the working hours for every nurse anesthetist, OR, holding and recovery nurse for the upcoming week, including what surgeries they assist.
|
|
11
|
Control planning
|
|
|
|
|
|
|
|
|
|
x
|
|
x
|
|
|
x
|
x
|
x
|
|
|
x
|
|
|
|
x
|
|
|
Check all requirements for the surgery to be able take place, determine the final order of the surgeries for each OR and if necessary revise planned surgeries.
|
|
12
|
Pick materials
|
|
|
|
|
|
|
|
|
|
x
|
|
x
|
x
|
X
|
|
|
|
|
|
|
|
|
|
|
|
|
Collect materials required for surgeries from the storage rooms and deliver these to the operating rooms.
|
|
13
|
Emergency admission
|
x
|
|
x
|
|
|
x
|
x
|
|
|
|
|
x
|
|
|
|
|
|
|
|
|
|
|
x
|
x
|
|
|
Define diagnosis and treatment for patients admitted to the Emergency Department and plan and prepare the patient for surgery.
|
|
14
|
Prepare patient on ward
|
|
X
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
x
|
|
x
|
Admit the patient to the nursing ward, administer premedication to the patient and further prepare the patient for surgery.
|
|
15
|
Prepare patient in holding
|
|
X
|
|
|
|
|
|
|
x
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
x
|
|
|
Transfer the patient from the nursing ward to holding, further prepare the patient for surgery and transfer the patient to the nurse anesthetist.
|
|
16
|
Make Radiology image
|
|
|
|
|
|
|
|
|
|
|
|
x
|
|
X
|
x
|
|
|
|
|
|
x
|
x
|
|
|
|
|
Ask the radiology department to make an image of the patient in a specific place (OR or recovery) and time.
|
|
17
|
Perform surgery
|
x
|
X
|
x
|
|
|
|
|
|
|
|
|
|
|
X
|
|
|
|
|
|
|
|
|
x
|
|
|
|
Perform the surgery with the OR team.
|
|
18
|
Clean OR
|
|
|
|
|
|
|
|
|
|
|
x
|
|
|
X
|
|
|
|
|
|
|
|
|
|
|
|
|
Ask the cleaning services to clean the operating room right after transferring the patient to recovery.
|
|
19
|
Order emergency CSD services
|
|
|
|
|
x
|
|
|
|
|
|
|
x
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Call the central sterilization department to ask for immediate cleaning and sterilization of medical instruments, as these may be re-used for surgery on the same day.
|
|
20
|
Patient care recovery
|
x
|
X
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
x
|
|
x
|
|
|
Take care of the patient after surgery, making sure the patient is well enough to be transferred to the nursing ward.
|
|
21
|
Aftercare of patient
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
x
|
x
|
|
x
|
Take care of the patient after surgery, making sure the patient is well enough to go home.
|
|
22
|
Manage OTC day program
|
x
|
X
|
|
|
|
|
|
|
x
|
x
|
|
x
|
|
X
|
|
x
|
x
|
|
|
|
|
x
|
x
|
|
|
|
Coordinate and manage the daily OR program, making sure that all surgeries planned for each day are well-executed and on time, and review this over time.
|
|
23
|
Manage OTC tasks
|
|
|
|
x
|
|
|
|
|
|
x
|
|
x
|
|
|
|
x
|
|
|
|
|
|
|
|
|
|
|
Coordinate and manage OTC operations over the long term.
|
|
|
Total number of tasks involved
|
8
|
7
|
2
|
2
|
1
|
1
|
1
|
1
|
4
|
8
|
1
|
11
|
1
|
8
|
2
|
5
|
7
|
1
|
1
|
3
|
1
|
5
|
8
|
6
|
2
|
3
|
|
- The entire social network
Figure 3 shows the entire social network with all agents and the ties between these agents. The names of all agents were abbreviated in the network figures and are explained in Additional File 3. Even though Figure 3 does not reveal the details of the network, it clearly shows that all agents are connected in one way or another and that there are no agents or cliques that are completely disconnected from the rest of the network. The relatively low density of 0.16, as shown in Table 3, indicates that there are agents or groups which are less connected, suggesting differentiation. The high number of cliques also indicates the presence of subsystems, demonstrating differentiation. However, 65% of all agents are part of multiple cliques across two related tasks. This high multiplexity value implies that there is integration as well. The spread between average and highest values for degree and betweenness centrality suggest that a relatively small number agents play an integrative role.
Table 3 Network metrics overall network
Network Parameter
|
Value
|
Number of agents
|
635
|
Number of ties
|
31499
|
Density
|
0.16
|
Number of cliques
|
8698
|
Multiplexity
|
413/635 = 65%
|
Agent parameter
|
Lowest
|
Average
|
Highest
|
Degree
|
1
|
99
|
399
|
Betweenness centrality
|
0
|
347
|
31379
|
Figure 3 also shows groups of agents who are closely connected, which suggests the presence of subsystems. We see groups of agents who share the same task or knowledge, or they deal with a specific patient group depending on age, condition or required length of stay. Examples of agents sharing the same task and patient group are on the top side edges of the network where we see the ER nurses and on the right side the nursing wards, which are all cliques; clockwise the groups of nurses are visible with codes KDVNUR, N2NUR, B0NUR, N1NUR, N0NUR, B2NUR, A2NUR. Each code starts with the name of the nursing department as defined by Slingeland Hospital, e.g., KDVNUR1, KDVNUR2 et cetera are nurses from department KDV. They also form subsystems because these nurses are all involved in the same task. Interestingly, the team leaders (WTEAM) of nursing wards B0, N1, N0, B2, IC have fewer connections to others in the hospital in comparison with the nurses, illustrated by their peripheral position in the network.
The group of intensive care (IC) nurses (ICNUR) form a clique as well, but they are more centrally positioned. This is because IC nurses have connections to all other nursing wards, as IC patients are always transferred to another nursing ward before they are discharged. Agents working in the daycare department F2, where patients stay because of their expected one-day length of stay, are also more centrally located because patients are transferred in case they need to stay the night.
All OR nurses (ORAS) also form a group, in three cliques, because they are divided into three clusters which are based on shared knowledge of medical disciplines. The holding and recovery nurses each have a clique as well. The anesthesiologists (AN) are visible as a group as well as the nurse anesthetists (ANNU), who are in the middle of the network. The surgeons do not form one group, but they form nine cliques that each share the knowledge of a specific medical discipline. Here we see separate subsystems according to medical discipline, which essentially all perform the same task. This is also the case for the secretaries of the outpatient departments, who are visible in the bottom left part of Figure 3.
The high number of cliques is largely explained by the fact that there are 7640 unique cliques that perform surgery (task 17). This will be analyzed further in the network analysis of each task.
The average degree is 99 and standard deviation is 79, which suggests centralization, as there are relatively large differences between the number of ties of agents. The agent with the highest degree is a nurse anesthetist with 399 ties to other agents. The nurse anesthetists all have a high degree, with an average of 387 ties. On the day of surgery they have interaction with all surgery team members, including surgeons, anesthesiologists and OR nurses. Furthermore, they interact with all ward nurses, and with holding and recovery nurses throughout the year. This is also the case for holding and recovery nurses who have an average degree of 300 and 318, respectively. The agents with relatively low degrees are on the edges of the network in Figure 3, e.g., all staff from the Central Sterilization Department (CSD) on the top left. In Additional File 3 the degrees of all agents are presented.
The OTC day coordinator (OTCO) has the highest betweenness centrality, which makes sense given the name of that function, but at the same time it is striking, because he does not contribute to multiplexity. The OTC capacity planner has the second highest centrality, and she has a strong integrative role between related tasks. The nurse anesthetists (ANNU) have high betweenness centrality as well as having a high degree, which also suggests a broker role.
The number of agents and communication links between them are different in the four time horizons which were presented in Figure 2. Table 4 clearly shows that the number of agents interacting and the density is higher on the day of surgery than before that day. If we look at the planning and execution phase, the density is 0.08 and 0.16 respectively. This suggests that, even though the overall network integration is low, in the months, weeks and days before surgery there is more differentiation and less integration than there is on the day of surgery. Furthermore, the OTC capacity planner plays a more prominent role before the day of surgery, whereas the OTC day coordinator is mainly involved on the day of surgery.
Table 4 Network metrics of the network over time
Time horizon
|
Number of agents
|
Number of ties
|
Density
|
Highest betweenness centrality
|
3-6 months
|
168
|
1041
|
0.07
|
OTPLAN
|
2-12 weeks
|
146
|
695
|
0.07
|
OTPLAN
|
1 day to 2 weeks
|
144
|
428
|
0.04
|
OTPLAN
|
Day of surgery
|
605
|
30135
|
0.16
|
OTCO
|
In the next section we will go into more detail of the network for each task.
- Network analysis per task
The social network per task is included in Figures 1 to 23 in Additional File 2. Table 5 shows the differences in network metrics between tasks. The number of participating agents varies from 4 to 391, the density from a low 0.01 to the maximum of 1, the number of cliques varies from zero to 7640 and clique overlap is between zero and 92%.
Table 5 Network metrics for each task
|
Tasks
|
Number of agents
|
Number of ties
|
Density
|
Number of cliques
|
Clique overlap
|
Organization unit
|
1
|
Make OR master schedule
|
28
|
110
|
0.3
|
2
|
6
|
21%
|
Cross functional
|
2
|
Make clinical bed plan
|
4
|
6
|
1.0
|
1
|
N/A
|
|
Cross functional
|
3
|
Schedule surgeons and anesthesiologists
|
70
|
206
|
0.09
|
10
|
0
|
0%
|
Cross functional
|
4
|
Schedule OTC nurses
|
88
|
801
|
0.2
|
6
|
1
|
1%
|
OTC
|
5
|
Plan equipment maintenance
|
4
|
6
|
1.0
|
1
|
N/A
|
|
Cross functional
|
6
|
Plan patient
|
92
|
315
|
0.08
|
48
|
1
|
1%
|
Cross functional
|
7
|
Request and order materials
|
61
|
139
|
0.08
|
0
|
N/A
|
|
Cross functional
|
8
|
Pre-operative screening
|
27
|
109
|
0.31
|
12
|
9
|
33%
|
Cross functional
|
9
|
Request and make appointment
|
56
|
140
|
0.09
|
0
|
N/A
|
|
Cross functional
|
10
|
Plan OTC nurses
|
85
|
84
|
0.02
|
0
|
N/A
|
|
OTC
|
11
|
Control planning
|
54
|
234
|
0.16
|
2
|
3
|
6%
|
Cross functional
|
12
|
Pick materials
|
55
|
107
|
0.07
|
53
|
2
|
4%
|
OTC
|
13
|
Emergency admission
|
139
|
2,840
|
0.30
|
1
|
N/A
|
|
Cross functional
|
14
|
Prepare patient on ward
|
314
|
11071
|
0.23
|
171
|
19
|
6%
|
Cross functional
|
15
|
Prepare patient on holding
|
289
|
1100
|
0.03
|
285
|
4
|
1%
|
Cross functional
|
16
|
Make radiology image
|
53
|
491
|
0.36
|
0
|
N/A
|
|
Cross functional
|
17
|
Perform surgery
|
148
|
5444
|
0.50
|
7640
|
136
|
92%
|
Cross functional
|
18
|
Clean OR
|
53
|
102
|
0.07
|
0
|
N/A
|
|
OTC
|
19
|
Order emergency CSD services
|
19
|
18
|
0.11
|
0
|
N/A
|
|
Cross functional
|
20
|
Patient care recovery
|
241
|
2355
|
0.08
|
285
|
10
|
4%
|
Cross functional
|
21
|
Aftercare of patient
|
391
|
12537
|
0.16
|
178
|
266
|
68%
|
Cross functional
|
22
|
Manage OTC day program
|
184
|
189
|
0.01
|
1
|
|
|
OTC
|
23
|
Manage OTC tasks
|
6
|
14
|
0.93
|
2
|
4
|
67%
|
OTC
|
Tasks with a relatively low density suggest differentiation. In Additional File 2 we see two network structures for such tasks: a network with weakly connected or disconnected cliques and a network with centralization. Task 3 (Figure 3 in Additional File 2) is a clear example of a network with ten cliques that are all disconnected. Here we see differentiation according to medical discipline with regard to how surgeons and anesthesiologists are scheduled. Each medical discipline represents a subsystem. This is also the case for task 6, but here the medical disciplines are situated around the OTC capacity planner in a star network.
Other tasks with a centralized network are 7, 9, 10, 12, 15, 18, 19, 20 and 22. The centralization is first explained by the fact that tasks are coordinated by the OTC day coordinator (tasks 7, 10, 19, 22). For the other tasks there is centralization because the central agent in each network interacts with each agent individually, while these agents do not interact with one another for that task. For example, on a regular basis the two logistical staff members ask all OR nurses, the OTC day coordinator and the OTC capacity planner for information on surgeries for which they pick the materials. Based on the definition of subsystems, these star networks do not have subsystems, because the agents are not highly connected.
Interestingly, the central agents in these star networks do not have a hierarchical position towards the agents around them, because the networks are cross functional (tasks 6, 7, 15, 19, 20). For tasks 4, 10, 12, 18 and 22 the central agents do not have a formal hierarchical position towards the other agents either.
For tasks with a higher density such as tasks 1, 2, 5, 8 and 11 we see integration, either by the presence of one clique (tasks 2 and 5) or multiple cliques (tasks 1, 8, 11). Furthermore, we see a network for task 14 with a highly connected group or subsystem with multiple agents in the center (Figure 14 in Additional File 2). Doing surgery in the OR (task 17) looks like a cloud of connections (Figure 17 in Additional File 2) because surgeons, anesthesiologists, OR nurses and nurse anesthetists work together in 7640 different cliques.
Besides density, clique overlap is an indication of integration. For tasks 1, 8, 17, 21 and 23 there is a relatively high overlap of 33% up to 92%, but for the other tasks clique overlap has a maximum of 6%, in which case the integration depends on just a few agents.
Remarkably, almost all tasks in which integration is observed are all organized in a cross functional manner. Managing the OTC tasks is the exception here, as this is done by agents who work only for the OTC department.
With regard to betweenness centrality per task, different agents act as a broker. The OTC capacity planner is most central for making the OR master schedule and planning surgeries. The OTC day coordinator is most central for scheduling and planning OTC nurses, for ordering materials and for responding to emergency orders from the CSD. For other tasks the agents with the highest betweenness centrality are two surgeons (task 3), the preoperative nurses and secretaries (task 8), the Neurology and Cardiology nurses (task 9), OTC logistical staff (task 12), the nurse anesthetists (task 14), holding nurses (task 15), one OR nurse (task 17), cleaning staff (task 18) and recovery nurses (task 20).
If we consider the four time periods of Figure 2 we see that the values for numbers of cliques and clique overlap are significantly higher on the day of surgery than for before that. This suggests that there is more differentiation as well as integration on the day of surgery than in the phases before. The high number of cliques is explained mainly by the fact that teams often interact for one specific patient; because these teams change frequently throughout the year, this results in a high clique overlap. In contrast, in the first phase, six to three months before surgery, there is a permanent smaller set of agents who make the OR master schedule, the clinical bed plan and equipment maintenance plan. Here we mainly see integration and no differentiation. At the same time the scheduling of surgeons and anesthesiologists (task 3) is executed by disconnected cliques, which shows differentiation.