RESEARCH AND PRACTICE

Social Network Diagramming as an Applied Tool for Public Health: Lessons Learned From an HCV Cluster Katarina M. Grande, MPH, Marisa Stanley, MPH, Carrie Redo, RN, BSN, Amy Wergin, RN, BSN, Sheila Guilfoyle, BA, and Mari Gasiorowicz, MA

Investigations of infectious disease clusters and response and control efforts can rapidly consume health department resources. Tools and strategies that can effectively focus interventions are needed. One example is social network diagramming and analysis, which was successfully used to continuously guide follow-up in a large HCV cluster in a rural Wisconsin county from 2010 to 2012. In outbreaks and clusters of infectious disease, social network analysis adds a layer to the traditional epidemiological approach that describes the course of an outbreak in terms of individual-level data. Incorporation of network properties illustrates underlying transmission dynamics based on social contacts, which can further define one’s “risk space.”1---3 The term “social network” refers to connections or relationships among individuals. The type of relationship depends on what the network is being used to examine, but the network can include friendships, sexual relationships, or partners with whom needles for injecting drugs are shared.4 In the context of public health, network analysis has been used to study disease transmission networks, social support and social capital networks, and organizational networks.5 In an applied sense, social network concepts have been incorporated into tuberculosis contact tracing procedures,6 as well as into some sexually transmitted disease cluster analyses and investigations.7,8 In 1 example, a health department redistributed the assignment of cases to disease investigators based on neighborhood so they could establish relationships with key influencers in the community, and thereby, increase the likelihood of changing behavior via these key “network nodes.”9 Nonetheless, social network approaches remain rarely used as on-the-ground tools in the context of clusters investigations.5 One way that network approaches can be immediately applicable is through network diagramming, which illustrates connections between entities

Objectives. We present an applied example of social network diagramming from 2010 to 2012 that was used to guide follow-up in a large HCV cluster in rural Wisconsin. Methods. In addition to collecting standard individual-level attributes, we also obtained partner-level information. Both sets of data were input into a network diagramming program to create a series of diagrams that emphasized variables, such as risk factors, key location in the network, and number of partners. Results. The visualization and cluster analysis guided testing and intervention priorities, were useful in sharing de-identified information about the cluster between health departments and community organizations and illustrated the key role young females played in holding the cluster together. Conclusions. Social network diagramming should be considered a practical and important public health tool for use in cluster management. (Am J Public Health. 2015;105:1611–1616. doi:10.2105/AJPH.2014.302193)

using shapes connected by lines. Visualizing the network in real time can help public health practitioners quickly identify individuals for follow-up, individuals who are critical in the spread of the infection, or individuals at high risk of becoming infected.10 This practical approach, recently described by Devakumar et al.,11 clarifies associations visually and allows for more guided discussion and planning by public health officials. We add a case example to this evidence base by describing the applied use of social network diagrams in a large HCV cluster investigation in a rural Wisconsin county.

CONTEXT OF THE CLUSTER INVESTIGATION The epidemiological profile of HCV infection in Wisconsin mirrors that of many states recently: clusters in rural areas among young people are increasing and associated with use of injection drugs.12---14 From 2003 to 2012, Wisconsin saw a decrease in the rate of reported HCV in the 30 to 49 years age group (58 to 32 per 100 000)15 alongside an increase in those younger than 30 years (15 to 54 cases per 100 000 persons).16 These changes coincided with an increase in newly reported cases

August 2015, Vol 105, No. 8 | American Journal of Public Health

in women—from 30% of all reported cases to 40% across the same time period (unpublished state surveillance data analyzed by authors; Wisconsin Department of Health Services, Division of Public Health, Hepatitis C Program’s National Electronic Disease Surveillance Systemcompliant database). Manitowoc County (population 81 44217 ) in rural, northeastern Wisconsin, saw a particularly steep increase in its reported cases of HCV infection, from an average of 18 cases per 100 000 from 2003 to 2009 to 52 from 2010 to 2012. Furthermore, in 2011, 41% of these patients were younger than 30 years.18 During the summer of 2010, the Manitowoc County Health Department received a report of an individual newly infected with HIV who would not provide the names of any contacts. Late in 2010, 2 anonymous HIV test clients named the aforementioned individual newly infected with HIV as a partner with whom they had shared injection drug use equipment. The anonymous test clients also named another individual with whom they had shared equipment who had recently tested positive for HCV at the health department. Throughout 2011 and 2012, the investigation grew to include a network of more than 240 cases or contacts that injected heroin and used a variety of

Grande et al. | Peer Reviewed | Research and Practice | 1611

RESEARCH AND PRACTICE

opiates and prescription drugs, such as morphine, codeine, Vicodin (hydrocodone/ acetaminophen), Percocet (oxycodone/ acetaminophen), Dilaudid (hydromorphone), oxymorphone, fentanyl, and methadone. The group also displayed behavioral risk factors of trading sex for drugs or money, having unprotected anal sex, and participating in “party groups” where sexual and drug encounters were communal and not well recollected. Although the majority of individuals with known ages in the group were younger than 30 years, a strong intergenerational component existed: 18 members were related, either as parents and children, or siblings. The local health department staffing included 1 public health nurse who shifted her duties to nearly full-time cluster response and management for 18 months, 1 manager who spent 20% of her time dedicated to oversight and resource management of the cluster, 2 to 3 public health nurses who provided temporary case management assistance at the outset of the cluster investigation, and 1 student worker who provided data entry support 1 day per week for 18 weeks. In addition, to manage the influx of data, the local health department partnered with the state health department to also temporarily assist with data management.

METHODS The Manitowoc County Health Department public health nurse team developed a standardized interview form on which to collect information about clients and their contacts. They collected information on both an individual or “attribute variable” level and a partner (social network) level, which denoted connections to others in the cluster (Figure 1). For each individual, they recorded demographic information and HCV, HIV, and tuberculosis testing status. For each partner named, they collected known demographic characteristic information and partner type data, that is, the type of partnership the 2 shared. Specific partnership types were (1) sexual contact (oral, anal, vaginal), (2) person who injects drugs (PWID; shared needle, cottons, water, or “parallel” [a term developed to describe individuals who used injection drugs in the same vicinity as a named partner, but unsure whether needles or other equipment

Attribute data Variable 1 (Age) Variable 2 (Status) Variable 3 (Gender) Person X 31 Positive M Person Y 19 Negative M Person Z 25 Positive F Descriptive variables linked to individuals such as HCV status, age, and gender Partner data Person X Person Y Person Z Person X 0 1 1 Person Y 1 0 1 Person Z 0 1 0 Partner linkages are read by the program from row to column—that is, Person X named Person Y and Person Z as contacts; Person Y named Person X and Person Z as contacts, and Person Z named Person Y as a contact. An illustrative network diagram of this partner data is below.

X

Y

Z Note. Squares represent males, circles represent females; gray shading represents positive status, no shading represents negative status; dashed outline represents age younger than 30 years, solid outline represents age 30 years or older.

FIGURE 1—Attribute data vs partner data example.

were shared]), and (3) both sexual and PWID partnership. The nurses interviewed each client in person upon testing to collect additional information on sexual risk and drug use. Clients were tested for HIV, tuberculosis, and hepatitis A, B, and C viruses.

Network Diagramming We developed a hand-drawn social network diagram on a ream of paper 15 by 3 feet (data available as a supplement to the online version of this article at http://www.ajph.org). We recorded new partnership information manually on the paper each time it was collected. The visualization of the cluster assisted in intervention efforts—the public health practitioners could rapidly see which individuals in the cluster were connected and carefully guide their testing and intervention priorities based on the connections. The approach of manually drawing the network diagram, however, quickly became cumbersome and functionally limited as the size of the cluster increased.

1612 | Research and Practice | Peer Reviewed | Grande et al.

To transfer the paper diagram to a computer, state health department employees developed an Access (Microsoft, Redmond, WA) database to keep track of testing and risk data. From this database, we exported data to construct social network diagrams using UCINet and accompanying visualization software, NETDRAW.19 We constructed network diagrams to depict connections between partners of all levels of partnership risk: sexual, injection drug, and both. We incorporated variables of age, gender, and HCV testing status into the diagrams by changing the shape, size, and color of the nodes. Multiple versions were created, including the entire network, focused diagrams featuring individuals with the greatest numbers of known partners, and specific subgraphs containing important connecting nodes. We added newly available information to the database in real time, and updated diagrams were generated. To ensure confidentiality, diagrams containing client codes were shared between the health departments over a secure, encrypted e-mail system. De-identified diagrams were

American Journal of Public Health | August 2015, Vol 105, No. 8

RESEARCH AND PRACTICE

produced alongside to illustrate cluster characteristics to a wider audience.

Network Metrics We developed diagrams including the social network factors. We used 3 main metrics to guide the diagrams: degree centrality, k-core, and betweenness centrality. Degree centrality is the number of connections each node had.20,21 Emphasizing nodes with high degree centrality displayed the individuals with the largest number of partners—either reported by the individual or the partner(s) of that individual. The k-core method describes subnetworks within the entire network and categorizes individuals into a group depending on the number of connections they have.8,22 Categorizing individuals into k-core groups can denote which subgroups of people are most connected to one another, which can better describe the social structure of the groups of acquaintances in the cluster. From a follow-up standpoint, k-core could be a useful metric for understanding how to potentially access key group members. The metric of “betweenness centrality” depicts which actors in the network were essential connectors to other actors—that is, if those with high betweenness centrality were removed from the network, a pathway between large groups of individuals would frequently no longer exist.22---24 Emphasizing those in the network with high betweenness centrality could indicate pathways potentially important for the spread of HCV. For optimal visualization, we used “springembedding” as a layout format. Springembedding is based on an algorithm that accounts for the distance between 2 nodes in a diagram; it allows closer actors to be pulled together and more distant actors to be pulled apart until a visualization is produced that more accurately represents the distribution of relationships.25

RESULTS The cluster included 243 individuals comprised of cases and their partners, including individuals who had not been tested and interviewed. As presented in Table 1, cluster membership was predominately male; of the 243 individuals, 131 (53.9%) were male,

78 (32.1%) were female, and 34 (14.0%) were unknown. (This was potentially caused by limited recollection of partner details during parties where injection drugs were shared, or the lack of gender-specific information recorded by public health staff on partners within this type of nonspecific large group setting because of an overall lack of identifying information that could lead to testing.) We listed these unknown individuals as contacts, but they appeared on the periphery of the network and were often impossible to identify to test. Most of the members whose age was known were younger than 30 years with an age distribution of 63 (25.9%) members who were younger than 25 years, 54 (22.2%) who were between 25 and 30 years, and 86 (35.4%) who were 30 years old or older. The overall median age was 29 years. The average number of connections within the 3 groups with known age was similar across groups: those who were younger than 25 years had a mean of 4.0 named connections; those who were ages 25 to 30 years had a mean of 3.8 connections, and those who were 30 years or older had a mean of 3.9 connections. Overall, 20 (8.2%) individuals had at least 5 connections, 8 (3.3%) had at least 4, 42 (17.3%) had at least 3, 48 (19.8%) had at least 2, and 125 (51.4%) had at least 1 connection. Of the 115 individuals tested for HCV, 80 (69.6%) tested positive. PWID was the most common risk factor, with 189 (77.8%) individuals identifying this as a risk factor. Regularly updated diagrams and information were provided to the Manitowoc County Health Department. The most connected individuals were of particular interest to us, and thus, a list of individuals with the largest number of connections was provided along with diagrams containing identifiers (Figure 2, de-identified). The most connected individuals were not immediately obvious before network diagrams were generated because often individuals would not name all the persons that named them—that is, on the network diagram a given arrow could be monodirectional rather than bidirectional. Furthermore, this diagram allowed us to quickly identify highly connected individuals with unknown HCV status (Figure 2) and highly connected individuals with negative HCV status (Figure 2). Highly connected individuals within both groups were a priority for further follow-up and intervention. Diagrams and subsequent lists were generated of

August 2015, Vol 105, No. 8 | American Journal of Public Health

TABLE 1—Descriptive Statistics of Cluster (n = 243): Manitowoc County, WI, 2010–2012 Variable

No. (%) or Median

Total

243 (100)

Gender Female

78 (32.1)

Male Unknown

131 (53.9) 34 (14.0)

Median age, y

29

Female

28

Male

30

Age, y < 25

63 (25.9)

25–30

54 (22.2)

> 30 Unknown

86 (35.4) 40 (16.5)

HCV statusa Positive

80b (69.6)

Negative

35b (30.4) Risk factor

PWID partner Yes No Unknown

189 (77.8) 54 (22.2) 0 (0.0)

Network properties: k-core membershipc 1-core

125 (51.4)

2-core

48 (19.8)

3-core

42 (17.3)

4-core

8 (3.3)

5-core

20 (8.2)

Note. PWID = person who injects drugs. a There were 128 persons with unknown HCV status who were not included in the percentage calculation. b Status among those tested. c K-core describes a group of people all of whom have at least k connections and are connected to at least k other people in the group. Membership of a 5-core (the highest risk group), for example, means the person has 5 connections, each of whom also have 5 connections to that group.

the most connected individuals within the entire network, as well as the most connected individuals within each status group of HCVpositive, HCV-negative, and HCV-status unknown. The public health response involved targeting the HCV-negative individuals who were highly connected to positive individuals within the network, as well as getting the highly connected unknowns tested. Linking those who

Grande et al. | Peer Reviewed | Research and Practice | 1613

RESEARCH AND PRACTICE

DISCUSSION

a

b

c

d

HCV positive HCV negative HCV status unknown Female

Note. This series depicts the 4 most connected individuals in the network (starred). As these individuals’ partners are diagrammed in sequence, connections within the core increase in overlap—that is, the partners of these 4 individuals are highly connected to one another. This visualization can quickly communicate to public health practitioners which individuals may be targeted for enhanced follow-up, such as the highly connected male in panel d with unknown status (denoted with “A”) and the highly connected female with negative HCV status (denoted with “B”).

FIGURE 2—Network diagram of 4 most connected individuals and their contacts showing (a) the most connected individual in the network (n = 43), (b) addition of the second-most connected individual and that person’s connections (n = 57), (c) addition of the third-most connected individual and that person’s connections (n = 65), and (d) addition of the fourthmost connected individual and that person’s connections (n = 71): Manitowoc County, WI, 2010–2012.

tested positive for HCV to services was also prioritized. Diagrams also provided a conceptual “big picture” of the cluster dynamics. For example, although the descriptive statistics indicated this cluster was mostly males, the network diagram illustrates how women were situated in very key positions in terms of potential for

promoting the spread of HCV (Figure 3, women had high betweenness). This universal view was also important in presenting the magnitude of the cluster to local policymakers such as the board of health, which made budgeting decisions, as well as discussing risk within the community (e.g., in high school).

1614 | Research and Practice | Peer Reviewed | Grande et al.

Social network diagrams provided an easily updatable visual diagram of the demographics characteristics and connections in 243 individuals included in our cluster investigation. The diagrams provided important information about the relational dynamics of the cluster and aided public health officials in directing prevention and testing efforts. Follow-up priority for testing and counseling was based on the level of risk, both at an individual level and partner level. For example, if an individual had only 1 partner, but that partner was highly involved with key members of the network, the individual might warrant more urgent followup. In addition, such visualization could promote less loss to follow-up, as health staff could instantly see which individuals were still in need of testing. Subsets of the cluster based on differing risk or demographic characteristics information could rapidly be produced as well; once the data were input into the system, generating multiple diagrams that contained different information took very little time. For example, 4 individuals in the network were coinfected with HIV, and generating their network diagrams could be useful for tailored prevention efforts. Because risk information related to sexual behavior was also collected, network diagrams focused only on sexual risk factors could be generated to produce a different risk picture. In addition to providing risk stratification detail, the diagrams also served the important purpose of sharing information between the local and state health departments. A de-identified network diagram could succinctly and quickly describe the magnitude of a cluster, which proved useful when external staff were called in to aid in response efforts. Network diagrams such as these could be used to convey information to agencies or funders beyond the health department, because these diagrams are easily understood.5

Limitations Social network analysis used outside—as in this case example—of a very controlled, experimental context had its challenges. The assumptions of independent observations that serve as the foundation of statistical analyses do not necessarily hold true in social networks.

American Journal of Public Health | August 2015, Vol 105, No. 8

RESEARCH AND PRACTICE

HCV positive HCV negative HCV status unknown Female Gender unknown

FIGURE 3—Network diagram of entire network with nodes sized based on betweenness—the positioning most likely to hold subsets of the network together; larger nodes are key connectors: Manitowoc County, WI, 2010–2012. Each member was highly related to other members of the population. Boundaries of network populations were not enumerated before sampling. To address this limitation, we limited our use of social network analysis and instead focused on the visualization benefit of social network diagramming. As public health enters the realm of social network analysis, scholars from the network analysis field caution against the use of network analysis without a comprehensive understanding of the theory.23,26 Nonetheless, metrics such as k-core and betweenness sampled in this example could hold future value, in particular, in linking social network analysis to public health practice. The k-core method elucidated individuals in highrisk subclusters, and the betweenness factor identified key transmission bridges within the overall cluster. By using the computer program featured in this case example, both metrics were simple pointand-click calculations after data was inputted.

Another limitation of our approach was the somewhat major level of effort required to both set up the data entry in a format that was easily inputted into UCINet and to learn the basic functions of the program. In this case example, we input data into Access, exported it to Excel (Microsoft, Redmond, WA) in a way that could be read by UCINet, and visualized the data using NETDRAW. We had an analyst with expertise in Access, a fellow with time to learn the new program, and a colleague skilled in Excel to ensure proper coding was used to move the data from Access to Excel. Although our team composition certainly aided the process of quickly transforming data into diagrams, it is not essential to have such scopes of expertise. Data can be simply entered into 2 Excel spreadsheets: 1 listing partners and 1 listing individual attribute data. If done in the format presented in Figure 1, no further data formatting need occur to paste the data into

August 2015, Vol 105, No. 8 | American Journal of Public Health

UCINet. Once the data are in UCINet, a number of online tutorials can be used to quickly learn the basics of network diagramming.27,28 In addition, the UCINet/NETDRAW software package was not free. It can be downloaded from the Internet at a cost of $40 for students, and $150 for government institutions.19 Free options that can conduct similar analyses and create diagrams are available online.29 UCINet was used in this case example because of our familiarity with it and its use in academic research. A cost estimate for the overall cluster investigation, including the network diagramming, was difficult to approximate because the surge of capacity to address the cluster was typical in public health investigations of clusters or outbreaks. The addition of the network diagramming aspect would take less staff time in the future because a system was established. A final limitation related to data quality. Because of the sensitive nature of the interview content and the number of individuals with incomplete risk or demographic characteristic information remaining in the diagrams, it was clear that some of the individuals interviewed did not provide comprehensive partner and risk information. However, because we were in charge of interviews, we developed trusting, nonjudgmental relationships with cluster members, thereby increasing the likelihood of collecting valid interview data. As of June 2014, we were still identifying additional people newly linked to the cluster; however, these individuals were not part of the cluster from 2011 to 2012 at the time of this case example. Because such newly linked individuals are currently being identified, but not additional individuals who were part of the cluster from 2011 to 2012, this further supported the likelihood that we captured most of the well-connected individuals in the network. Nonetheless, there were still individuals who refused interviews or could not be located, and were therefore not included in the diagram. Because the network properties were dependent on the characteristics of all of the partnerships in the entire network, it was plausible that the network could look different if every piece of information was collected.

Conclusions Because of the key positioning of women in this network diagram, it would be useful to compare the network location of women within clusters of similar demographic characteristics to understand if the social positioning was

Grande et al. | Peer Reviewed | Research and Practice | 1615

RESEARCH AND PRACTICE

related to higher likelihood of HCV spread, although directionality of HCV transmission will always remain a challenge with this type of analysis. Previous evidence suggests young female PWIDs are more likely to engage in higher risk injection practices, such as borrowing needles, sharing ancillary equipment, and being injected by someone else.30 As the previously male-dominated gender gap between newly HCV-infected young persons31 closes, it is particularly important to examine the changing dynamics of young female PWIDs and their social networks.32 This cluster example provided a case for the use of network diagramming as a tool to use alongside established contact tracing efforts, because it illustrated and quantified connections that were not well described by traditional methods. j

About the Authors At the time of the study, Katarina M. Grande was with the Population Health Institute, University of Wisconsin, Milwaukee. At the time of the study, Marisa Stanley was with the Wisconsin Department of Health Services, Division of Public Health, Madison. Carrie Redo and Amy Wergin are with the Manitowoc County Health Department, WI. Sheila Guilfoyle and Mari Gasiorowicz are with the Wisconsin Department of Health Services, Division of Public Health, Madison. Correspondence should be sent to Katarina M. Grande, Wisconsin Department of Health Services, Division of Public Health, AIDS/HIV Program, Room 230, 1 West Wilson Street, Madison, WI 53703 (e-mail: Katarina. [email protected]). Reprints can be ordered at http://www.ajph.org by clicking the “Reprints” link. This article was accepted July 3, 2014.

Contributors

Human Participant Protection No human participants were involved in this study. Because the purpose of this exercise was for public health response, rather than research, institutional review board approval was not warranted.

References 1. Stattner E, Vidot N. Social network analysis in epidemiology: current trends and perspectives. In: 2011 Fifth International Conference on Research Challenges in Information Science. Gosier, Guadeloupe; May 19-- 21, 2011; 1-- 11. 2. Rothenberg RB, Potterat JJ, Woodhouse DE, et al. Choosing a centrality measure: epidemiologic correlates in the Colorado Springs study of social networks. Soc Networks. 1995;17(3---4):273---297. 3. Ogilvie G, Knowles L, Wong E, et al. Incorporating a social networking approach to enhance contact tracing in a heterosexual outbreak of syphilis. Sex Transm Infect. 2005;81(2):124---127. 4. Klovdahl AS, Potterat JJ, Woodhouse DE, et al. Social networks and infectious disease: the Colorado Springs Study. Soc Sci Med. 1994;38(1):79---88. 5. Luke DA, Harris JK. Network analysis in public health: history, methods, and applications. Annu Rev Public Health. 2007;28:69---93. 6. National Tuberculosis Controllers Association; Centers for Disease Control and Prevention. Guidelines for the investigation of contacts of persons with infectious tuberculosis. Recommendations from the National Tuberculosis Controllers Association and CDC. MMWR Recomm Rep. 2005;54(15):1---47. 7. Khan MR, Epperson MW, Mateu-Gelabert P, et al. Incarceration, sex with an STI- or HIV-infected partner, and infection with an STI or HIV in Bushwick, Brooklyn, NY: a social network perspective. Am J Public Health. 2011;101(6):1110---1117. 8. Rice E, Barman-Adhikari A, Milburn NG, Monro W. Position-specific HIV risk in a large network of homeless youths. Am J Public Health. 2012;102(1):141---147. 9. Rothenberg R, Narramore J. Commentary: the relevance of social network concepts to sexually transmitted disease control. Sex Transm Dis. 1996;23(1):24---29.

K. M. Grande conducted the social network diagramming and led the writing of the article. M. Stanley provided epidemiological input and led the input of attribute and partner characteristics into the Access database. C. Redo and A. Wergin led the public health response, provided case interview data, and provided regular feedback on the applicability of the diagrams. S. Guilfoyle provided input into epidemiological characteristics of the cluster. M. Gasiorowicz provided epidemiological input, advice on data interpretation, and overall mentorship. All authors provided input into the content, organization, and structure of the article.

10. Rothenberg RB, Sterk C, Toomey KE, et al. Using social network and ethnographic tools to evaluate syphilis transmission. Sex Transm Dis. 1998;25(3):154---160.

Acknowledgments

13. Centers for Disease Control and Prevention. Hepatitis C virus infections among adolescents and young adults—Massachusetts, 2002---2009. MMWR Morb Mortal Wkly Rep. 2011;60(17):537---541.

K. M. Grande was a University of Wisconsin Population Health Service Fellow during this article’s development; the UW Population Health Service Fellowship is supported by funding from the Wisconsin Partnership Program. We wish to thank Lauren Stockman, Jim Vergeront (both from Wisconsin Department of Health Services, Division of Public Health), and Alexander Gutfraind (University of Illinois-Chicago) for their input and review of the article.

11. Devakumar D, Kitching A, Zenner D, et al. Tracking sickness through social networks—the practical use of social network mapping in supporting the management of an E. coli 0157 outbreak in a primary school in London. Epidemiol Infect. 2013;141(10):2022---2030. 12. Centers for Disease Control and Prevention. Notes from the field: hepatitis C virus infection among young adults—rural Wisconsin, 2010. MMWR Morb Mortal Wkly Rep. 2012;61(19):358.

14. Holmberg SD. Emerging epidemic of hepatitis C in young nonurban injection drug users (IDU). Report on Technical Consultation. Hepatitis C Virus Infection in Young Persons who Inject Drugs. Washington, DC: Department of Health and Human Services; February 26-27, 2013. Available at: http://aids.gov/pdf/hcv-and-youngpwid-consultation-report.pdf. Accessed May 6, 2014.

1616 | Research and Practice | Peer Reviewed | Grande et al.

15. Stockman LJ, Guilfoyle SM. Wisconsin hepatitis C (HCV) surveillance summary: cases reported 1/1/2012 through 12/31/2012. Available at: https://www.dhs. wisconsin.gov/publications/p0/p00440-2012.pdf. Accessed February 5, 2015. 16. Stockman LJ, Guilfoyle SM, Benoit AL, et al. Rapid hepatitis C testing among persons at increased risk for infection. MMWR Morb Mortal Wkly Rep. 2014;63(14): 309---311. 17. US Census Bureau. 2010. 2010 Population Finder. Available at: http://www.census.gov/popfinder/? fl=55001:55071. Accessed May 6, 2014. 18. Gasiorowicz M, Guilfoyle S, Stanley M, et al. Hepatitis C surveillance and cluster investigations in Wisconsin. Wisconsin AIDS/HIV Program Notes. 2011;1---5. Available at: https://www.dhs.wisconsin.gov/publications/p0/p0079211-november.pdf. Accessed February 5, 2015. 19. Borgatti SP, Everett MG, Freeman LC. UCINET for Windows: Software for Social Network Analysis. Boston, MA: Analytic Technologies; 2002. 20. Krebs V. Social network analysis: A brief introduction. 2000-2011. Available at: http://www.orgnet.com/ sna.html. Accessed September 16, 2012. 21. Christley RM, Pinchbeck GL, Bowers RG, et al. Infection in social networks: using network analysis to identify high-risk individuals. Am J Epidemiol. 2005;162(10):1024-- 1031. 22. Parchman ML, Scoglio CM, Schumm P. Understanding the implementation of evidence-based care: a structural network approach. Implement Sci. 2011;6:14. 23. Hawe P, Webster C, Shiell A. A glossary of terms for navigating the field of social network analysis. J Epidemiol Community Health. 2004;58:971---975. 24. Hanneman RA, Riddle M. Introduction to social network methods: Chapter 10—centrality and power. Riverside, CA: University of California; 2005. Available at: http://faculty.ucr.edu/?hanneman/nettext/C10_Centrality. html#Betweenness. Accessed February 21, 2014. 25. Freeman LC. Visualizing social networks. Available at: http://www.cmu.edu/joss/content/articles/volume1/ Freeman.html. Accessed September 15, 2012. 26. Butts CT. Revisiting the foundations of network analysis. Science. 2009;325(5939):414---416. 27. Hanneman RA, Riddle M. Introduction to Social Network Methods. Riverside, CA: University of California; 2005. 28. Hanneman RA, Riddle M. Introduction to Social Network Methods: Working With NetDraw to Visualize Graphs. Riverside, CA: University of California; 2005. 29. Huisman M, van Duijn MAJ. A reader’s guide to SNA software. In: Scott J, Carrington PJ, eds. The SAGE Handbook of Social Network Analysis. London, England: Sage; 2011:578---600. 30. Evans JL, Hahn JA, Page-Shafer K, et al. Gender differences in sexual and injection risk behavior among active young injection drug users in San Francisco (the UFO study). J Urban Health. 2003;80(1):137---146. 31. National Alliance of State and Territorial AIDS Directors. Hepatitis C and young people who inject drugs. 2014. Available at: https://www.nastad.org/docs/Fact% 20Sheet%20NASTAD%20YPWID%20and%20HCV. pdf. Accessed June 26, 2014. 32. Page K, Morris MD, Hahn JA, et al. Injection drug use and hepatitis C virus infection in young adult injectors: using evidence to inform comprehensive prevention. Clin Infect Dis. 2013;57(suppl 2):S32---S38.

American Journal of Public Health | August 2015, Vol 105, No. 8

Social Network Diagramming as an Applied Tool for Public Health: Lessons Learned From an HCV Cluster.

We present an applied example of social network diagramming from 2010 to 2012 that was used to guide follow-up in a large HCV cluster in rural Wiscons...
861KB Sizes 0 Downloads 9 Views