Background

In Australia, the Australian Health Protection Committee (AHPC) is chaired at Deputy Secretary level by the Australian Government Department of Health and Ageing, and its core includes the chief medical officer; each state and territory chief health officer (CHO), and, if required, public health or emergency personnel nominated by the CHO for relevant agenda items; health disaster officials (up to 3) nominated by states and territories or the commonwealth; the chairs of 3subcommittees (Communicable Disease Network Australia, Public Health Laboratory Network, and the Environmental Health Committee; the director general of Emergency Management Australia; the head of Defence Health Services; a representative of the New Zealand Ministry of Health; a representative of Emergency Management Australia; representatives of the National Mental Health Disaster Response Committee; and others, as necessary (AHPC). ²

In the state of New South Wales (NSW), the Public Health Act designated the NSW Ministry of Health (NSWH) as the lead combat agency during the H1N1 2009 outbreak. NSWH had prepared plans to deal with different types of pandemics, including the human influenza pandemic; some of these plans are the State Disaster Plan, Human Influenza Pandemic Plan (HIPP), and NSW HEALTHPLAN. The HIPP was based on more severe types of influenza such as H5N1, whereas H1N1 was relatively mild.Reference Baber, Cross, Smith and Robinson ⁶

The Australian Health Management Plan for Pandemic Influenza categorizes 5 management phases: delay, contain, protect, sustain, and control, which were later collapsed to 3 phases: delay, contain, and protect. At the conclusion of any event for which the plan was activated, these plans should be updated with the introduction of major structural, organizational, or legislative changes in NSW, or at least every 5 years.Reference Baber, Cross, Smith and Robinson ⁶ NSWH is responsible for the management effort, which starts with surveillance and follows through different components, which include some of the followingReference Baber, Cross, Smith and Robinson ⁶ ):

• ∙ Various laboratories notify confirmed cases of influenza;

• ∙ The Public Health Real-time Emergency Department Surveillance System, which monitors near real time for influenza-like illness from most emergency departments (EDs) in NSW;

• ∙ Public health units (PHUs) receive reports from clinicians or institutions of unusual cases;

• ∙ Sample of general practitioners (GPs) contribute data on influenza-like illness to sentinel surveillance systems; and

• ∙ The Australian government supplies absenteeism data.

During the H1N1 outbreak, NSWH implemented additional monitoring and surveillance measures, including

• ∙ Active public health follow-up of possible and confirmed cases of pandemic influenza;

• ∙ Border screening for influenza-like illness in travelers from affected regions; and

• ∙ Data provided through collaborative efforts of multiple hospitals (eg, national data on intensive care or pediatric admissions).

During the pandemic, NSW was divided into 8 local health districts, each with a PHU that managed communication locally with health facilities such as hospitals, laboratories, and community health centers. Those laboratories and hospitals in turn communicated with NSWH directly, especially with specialized units within them, such as intensive care units (ICUs). Different bodies regulated the GPs, while the GPs communicated with the local and, occasionally, federal health authorities.

Table 1 shows the reported cases of infection in Australia for the H1N1 2009 outbreak. As seen in Table 2, the Hunter New England area had the lowest transmission rate (0.9 per 100 000 population) and the lowest number of confirmed cases in the state of NSW (only 8), whereas the adjacent North Coast health area had a transmission rate more than twice as high (2 per 100 000). Table 2 also indicates that within the same state the variation in infection rates was about eightfold. Considering that the demographics and pathogen were the same, we believe that the management and coordinated response led by corresponding agencies in each individual state and area health service within the states were factors affecting the infection rates and needed further investigation.

Table 1

Confirmed Cases of H1N109 Infection in Australia

^a Figures are based on estimated residential populations on December 31, 2008.

^b Total includes all territories.

Table 2

Confirmed Cases of H1N1 2009 Infection Within New South Wales

^a Figures are based on estimated residential populations on December 31, 2008.

^b Discrepancy with the New South Wales total in Table 1 is due to differences in population projections.

Coordination Complexity in Disease Outbreaks

Managing disease outbreaks relies substantially on information collection, validation, sharing, and visualization.Reference Tschoegl, Below and Guha-Sapir ^{7 ,} Reference Hitchcock, Chamberlain, Van Wagoner, Inglesby and O'Toole ⁸ Consequently, a multiagency coordinating complex system consisting of cross-disciplinary public and private health professionals supported by advanced information systems needs to be established. This system can be described as infectious disease informatics, which Hitchcock and coworkers define as “an interdisciplinary research area that focuses on the design, implementation, and evaluation of advanced systems, techniques, and methods for managing infectious disease and epidemic outbreaks, ranging from prevention to surveillance and detection.”Reference Freeman ⁹

In a similar way, the organizations that interact during disease outbreaks represent a unique form of interorganizational coordination. They create a matrix of interdisciplinary agencies coordinating within certain time constraints (disease infectivity characteristics). To understand the complexity of such a system, it is necessary to view the overarching structure and process holistically.

Figure 1 illustrates the general scheme and provides a brief overview of these tasks. It shows the communication and workflow paths between different conceptual tasks during the pandemic. For example, one cluster (case definition, horizon scanning, and surveillance and detection) leads to diagnosis, which then results in clinical management of confirmed cases to monitor the situation, and which then provides input to situation reports. Also resource management oversees vaccine preparation, which is used to manage reported cases.

Figure 1

Schematic Diagrams for the Tasks During the Outbreak.

Coordination structures are dynamic and change with the current crisis; hence, it is necessary to study the pandemic coordination before and during the outbreak. Usually pandemic management agencies participate in constant exchange of information such as details of new cases, confirmed laboratory results as part of surveillance, or horizon scanning activities. These activities are normally constant bureaucratic operations in which most public health-related agencies routinely engage as part of their standard practices. This information gathering and exchange process is an integral and important part of prepandemic administration that will also lead to a proactive management model. In theory, this communication should lead at least to increased management and coordination efforts during the outbreak itself, as the agencies are familiar with each other’s roles and communication protocols.

The present study was designed to capture the communication lines before and during the pandemic. This approach could facilitate understanding of the coordination lines and how they change with phase change. Moreover, coordination could be investigated more comprehensively during the outbreak. This opportunity allowed the process to be examined from both facets, which, to our knowledge, has not been undertaken previously. Figure 2 highlights the phases of coordination considered in this study that led to the proposed hypotheses. This report only provides a detailed analysis of formal/hierarchical coordination for preparedness and response to the H1N1 outbreak in Australia.

Figure 2

The 3 Phases of Coordination Considered in the Research.

Social Network measures for coordination

We use the following measures of social networks to assess the networks of preparedness and response to the Australian H1N1 outbreak.

Degree Centrality Measure

The number of links to and from a node is the most frequently used, intuitive, and easy to understand centrality measure. Degree centrality is a local centrality measure because it can be calculated without reference to the overall structure of the network. In an asymmetric (directed) network, in-degree is defined as the number of ties received and out-degree is the number of ties sent or initiated by the node outwardly. The following formulas define these types of vectored centrality:

(1.1) $$C_{{D_{i} }} (out{\minus}degree)=\mathop{\sum}{_{j} } d_{i} _{j} $$

(1.2) $$C_{{D_{i} }} (in{\minus}degree)=\mathop{\sum}{_{j} } d_{{ji}} $$

Where d _ji indicates the existence or non-existence of a link between nodes i and j. If a link exists between node i and node j, d _ji =1 If there is no link, d _ji =0.

In-degree counts the number of times a node (organization) is needed or nominated by the others in the network. To make the degree centrality comparable between networks of different sizes, the count is divided by N-1, which is the maximum possible number of connections a particular node can have, hence a representation of the network size. This results in a normalized centrality measurement that varies from 0 to 1Reference Lin ¹⁰ :

(1.3) $$C_{{D_{i} }} (out{\minus}degree)={{\sum_j d_{{ij}} } \over {N{\minus}1}}$$

(1.4) $$C_{{D_{i} }} (in{\minus}degree)={{\sum_j d_{{ji}} } \over {N{\minus}1}}$$

where N represents the number of nodes.

The significance of centrality arises when it is realized that organizations rely on strong ties to provide needed resources.Reference Burt ¹¹ Previous research has claimed that the ego-network measure—centrality—is associated with performance.Reference Rosenthal ^{12 ,} Reference Bavelas ¹³ Centrality depicts the degree to which an individual obtains information and controls benefits from non-redundant ties, and it is theorized to positively influence performance. Such assertions also are supported by the controlled experiment conducted by Bavelas in which individuals were given certain information to assemble to reach a decision, with the purpose of issuing orders to solve a puzzle.Reference Leavitt ¹⁴ Bavelas found that centralized structures (eg, star) performed better than decentralized ones (eg, circle). It was suggested that the reason was that decentralized information floats better than centralized information.

Numerous follow-up studies to the Bavelas experiment were conducted to assess the impact of different structures on communication and problem-solving performance. Unfortunately, these empirical studies did not produce consistent and cumulative results. In fact, some results were contradictory with each other and with the original Bavelas findings. For example, several researchers confirmed that the star structure performed better than the all-channel structure.Reference Mulder ^{15 –} Reference Guetzkow and Dill ¹⁸ On the other hand, other empirical research determined that the all-channel structure performed better than the star for simple task design,Reference Shaw ¹⁹ while still others found no significant difference between the performance of the star and all-channel structures.Reference Mulder ^{20 ,} Reference Freeman ²¹ Therefore, empirical and experimental results suggested that centrality can have positive, negative, or no effect at all on performance.

Closeness and Betweenness Centrality Measures

Closeness measures the average distance of a node from all other nodes in the network. Totaling these distances and then inverting the resulting value changes the measure from a distance measure to a closeness measure. Point closeness is then the inverted sum of all the distances, and normalized closeness is N-1 divided by the sum of distances, making it an average closeness measure. Normalized closeness is calculated as followsReference Freeman ^{21 ,} Reference Valente ²² ):

(1.5) $$C_{{C=}} {{N{\minus}1} \over {\sum^ \,D_{{ij}} }}$$

In-closeness refers to the links directed to a person and out-closeness refers to the links initiated from a person. The practical implication is that the person with the highest out-closeness value is the person who can reach others in the fewest number of steps, whereas the person with the highest in-closeness value is the person others can reach in the fewest number of steps.Reference Valente ²²

Betweenness centrality proposed by Freeman measures the frequency with which a person lies on the shortest path connecting everyone else in the network.Reference Freeman ²¹ The concept of betweenness is very appealing, as it measures the degree to which a node occupies a strategic position in a network, somewhat akin to a combination of bridging and centrality. Normalized betweenness centrality is calculated as follows:

(1.6) $$C_{b} = {{{{g_{{ij }} P_{k} } \over {g_{{ij}} }}} \over {n^{2} {\minus}3n{\plus}2}}$$

where g _ij P _k counts the number of times point k lies on the geodesic (shortest path) connecting all other nodes (i and j), and g _ij is the number of geodesics in the network. The maximum possible value that the numerator $$g_{{ij{\rm }}} {{P_{k} } \over {g_{{ij}} }}{\rm }$$ can reach is n ²−3n+2, so this is the normalization factor.Reference Freeman ²¹ ) Because a geodesic path is directional, betweenness centrality is directional and hence separate calculations for in and out directions are needed. Freeman explained that betweenness centrality captures a gate-keeping function: if members high in betweenness oppose an idea, its diffusion to other segments of the group might be blocked.Reference Freeman ²¹ Closeness centrality captures a communication role such that people high in closeness can rapidly communicate an idea to many others.Reference Valente ²²

Network Density

Density is the first measurement to be discussed. It is the number of connections in a network reported as a fraction of the total links, and is calculated as follows:

(1.7) $$D={l \over {N(N{\minus}1)}}$$

where l is the number of links in the network and n is the network size. Equation 1.7 is applicable to asymmetric networks, but the numerator must be multiplied by 2 for undirected (symmetric) networks. The relationship between size and density are inverse: as size increases, density decreases. One reason for this is that the number of other people a person knows or can establish relationships with has practical limits.

Data Collection and Analysis

Interviews with professionals were used to construct the network of the participating organizations in pandemic intervention and coordination. Interviews snowballed from 1 participant to that participant’s interesting alternatives who were considered to have knowledge that could further extend the scope of information. Table 3 shows some of the people in those positions who participated in the survey.

Table 3

Some of the Positions That Participated in Pandemic Coordination

After a detailed examination of the research that has been conducted in disease outbreak coordination, a gap in the research identified some interesting questions yet to be answered:

• ∙ How does one identify the initiation point for this multiagent coordination process and when it will be over?

• ∙ What are the characteristics of the organizations that will play a central role during the evolution of coordination?

The next step was to dissect the research questions into tangible ones, which were allotted focal points (Table 4).

Table 4

Summary of the Main Focal Points and Their Relevant Questions

Administering the Questionnaire

Table 5 shows the proposed matrix for each section of the questions, along with the proposed interviewees—the titles have been generalized to reflect different health authorities’ structures and names that might vary among states or countries.

Table 5

Matrix of Proposed Interviewees for the Qualitative Questionnaire

Administering the Survey

The egocentric network approach was more practical and less expensive, and hence was adopted for the study. It also allowed the participant to freely recall alternatives and could include alternatives not conceived of by the researcher. This response would not have been possible with the sociocentric approach. Table 6 shows the matrix of expertise targeted.

Table 6

Overview of the Positions Surveyed

The sociocentric method was used, based on a snowballing technique, to select the candidates for interview. It was decided to produce the data collection instrument in 2 phases to ensure that it was reliable and valid. An exploratory questionnaire was first designed and was used to further the design of the main instrument. The main instrument collected both qualitative and quantitative data. Quantitative data was used to provide empirical validation or rejection of the proposed hypotheses, and the qualitative data provided in-depth views of causes, reasons, and context for the empirical results. Surveys were conducted with approximately 70 respondents who represented a wide range of professional and bureaucratic backgrounds and positions within the network that managed pandemic intervention. This combination of the sociocentric relational network, the qualitative and quantitative instruments, and the respondents interviewed created a wealth of domain-related information about coordination related to the pandemic, as well as statistical data for validation.