Understanding the social and spatial relationships that connect people is a key element to stop the spread of infectious diseases. These networks are particularly relevant to combat epidemics among populations that are hard to reach with public health interventions. Network-based approaches, for example, can help to stop HIV or hepatitis C from spreading amongst populations that use injectable drugs. Yet how social and geographic connections such as acquaintances, injection partners, or preferred drug use places impact the risk of infection is still poorly mapped out.

To address this question, Clipman et al. focused on people who inject drugs in New Delhi, India, a population heavily impacted by HIV and hepatitis C. Over 2500 people were recruited, each participant inviting their injection partners to also take part. The volunteers answered survey questions, including where they used drugs, and provided a blood sample to be tested.

The results showed that, even after adjusting for individual risk factors, where people used drugs and with whom affected their risk of becoming infected with HIV and hepatitis C. In terms of social ties, the likelihood of HIV and hepatitis C infection decreased by about 13% for each person separating a given individual from an infected person. However, geographical networks also had a major impact. Injecting at a popular location respectively increased the odds of HIV and hepatitis C infection by 50% and 69%. In fact, even if the participant was not using drugs at these specific places, having an injection partner who did was enough to increase the risk for disease: for each person separating an individual from the location, the likelihood of being infected with HIV and hepatitis C decreased by respectively 14% and 10%.

The results by Clipman et al. highlight how the relationships between physical spaces and social networks contribute to the spread of dangerous diseases amongst people who inject drugs. Ultimately, this knowledge may help to shape better public health interventions that would take into account the importance of geographical locations.

People who inject drugs (PWID) bear a disproportionate burden of human immunodeficiency virus (HIV) and hepatitis C virus (HCV) infection and account for some of the fastest-growing epidemics globally. While there has been substantial progress in combating these epidemics, HIV and HCV prevalence and incidence among PWID remain high, especially in South, Southeast and Central Asia, and Eastern Europe (DeHovitz et al., 2014). Individual-level factors for infection are well established, but less is known about network and spatial drivers of HIV and HCV among PWID, especially from low- and middle-income settings.

Network-based interventions for HIV and HCV are increasingly being implemented; however, they are seldom informed by empirical sociometric network data and more often informed by models (Zelenev et al., 2018; Hellard et al., 2015; Rolls et al., 2013; Hellard et al., 2014). Existing data derive from small egocentric network studies of ‘indexes’ or ‘egos’ and their immediate connections (first degree ‘alters’) (Costenbader et al., 2006; Latkin et al., 2009; Latkin et al., 2011; Latkin et al., 2013; Latkin et al., 2010). Few studies have examined the broader sociometric network, which captures the alters of those first degree alters (second degree alters of the index), and so on, providing a more complete representation of the underlying network (Figure 1).

Figure 1

Download asset Open asset

Even less is known about the overlap of these egocentric and sociometric networks in space. While spatial heterogeneity of HIV/HCV burden has been previously described in high-income settings (Des Jarlais et al., 2018), less is known about whether transmission is driven more by injection partner connections versus spaces/venues people reside and/or inject within. Incorporating spatial data, specifically in the form of injection venues, can further inform whether independent sociometric injection networks overlap spatially to more comprehensively examine the distribution of HIV/HCV and assess the role of space in the diffusion of disease.

This manuscript aims to characterize egocentric, sociometric, and spatial network structures in a community-based sample of 2512 PWID in New Delhi, India and examine the role of individual- and network-level correlates of HIV and HCV infection.

In this sample of PWID in New Delhi, India, we observed an extremely high burden of HIV and HCV infection and strong associations between HIV and HCV within not only an individual’s immediate egocentric network but also broader sociometric, spatial, and sociospatial networks that incorporate indirect ties. These data are among the first to elucidate sociometric and spatial injection networks of PWID from a low- and middle-income country and provide critical insights into the design of HIV and HCV programming.

Empirical network data among PWID have often been limited to egocentric network data, which only capture information on individuals and direct contacts (Costenbader et al., 2006; Latkin et al., 2009; Latkin et al., 2011; Latkin et al., 2013; Latkin et al., 2010). These studies have shown that network instability or turnover in PWID’s injection partners promote HIV transmission. Limited sociometric data that exist come primarily from non-PWID populations in developed country settings and support the importance of understanding network connections beyond direct ties. For example, a seminal study assessing HIV and STI transmission in Colorado Springs found that HIV risk appeared low based on individual-level or egocentric network data, but sociometric data revealed risk to be higher than anticipated, with most individuals being within a few steps of an HIV infected person (Klovdahl et al., 1994; Rothenberg et al., 1995). Other studies in the United States have shown that sociometric networks can propagate HIV, with core individuals of large network components serving as centers of high-risk behavior and pockets of infection that could be targeted by network-based interventions (Friedman et al., 1997; Young et al., 2013).

The sociometric network presented here provides further support for transmission within large network components in a low- and middle-income country setting. For both prevalent HIV and active HCV infection, we found that while having direct ties with HIV/HCV infection was associated with prevalent HIV/HCV, sociometric factors such as network distance to an infected alter were also independently associated even after accounting for direct ties. Among HCV RNA positive persons, the average sociometric network distance to another HCV RNA positive participant was significantly shorter compared to those of anti-HCV positive persons with undetectable HCV RNA or anti-HCV negative persons, supporting that network proximity to PWID with HCV RNA infection indicates higher likelihood of reinfection.

These analyses further contribute to available network data by overlaying sociometric network data with information on injection venues. Prior studies, including some among PWID, have shown that physical spaces play an important role in the spread of disease, but have not examined spread through sociometric injection networks or associations with indirect connections to spaces (Zelenev et al., 2016; Gesink et al., 2014; Rudolph et al., 2017; Logan et al., 2016). The incorporation of space accounts for undocumented connections between participants in a space as well as spatial factors themselves (e.g., access to harm reduction services). Logan et al. demonstrated this in a sample of 600 participants (including 303 PWID) from Winnipeg, Canada, defining a geographically and socially cohesive community through which infections spread and identifying key venues driving such spread (Logan et al., 2016). This is one of the few studies that integrated space with injection network data but represents a small sample from a developed country setting where HIV burden was significantly lower and the social context of drug use is very different.

The idea that the sociospatial risk network may be the most relevant for transmission of disease is particularly salient with the increased focus on network-based interventions to reduce HIV and HCV transmission and optimize care outcomes while conserving resources. These findings suggest that in some settings, the sociospatial network may explain the majority of disease spread and interventions targeted at key spaces have the potential to interrupt transmission across a network and impact an entire city. For example, in this sample, it could be hypothesized that blanket coverage of venue #40 with treatment and pre-risk exposure prophylaxis, could impact transmission across New Delhi given the strong association of venue #40 with prevalent HIV among PWID in New Delhi.

Traditionally network-based interventions have relied heavily on social diffusion. For example, for HIV, network members are used to improve retention to antiretroviral therapy and improve viral suppression (Klovdahl et al., 1994), and ‘deep chain’ respondent-driven sampling is being used to identify undiagnosed or out-of-care HIV-positive men who have sex with men in the United States (Rothenberg et al., 1995). For HCV, egocentric network-based treatment approaches have been identified as the optimal approach to deliver therapy while minimizing reinfection (Rolls et al., 2013; Hellard et al., 2014). However, if networks are highly interconnected especially within other PWID at a venue as observed in New Delhi, failure to incorporate space in the consideration of a network could result in high rates of HCV reinfection.

In addition to network-level factors, these findings further reinforce the importance of well-established individual-level factors such as needle sharing, injection frequency, and experiencing homelessness. These associations and the limited uptake of harm reduction in this sample support continued efforts to expand harm reduction in this population. A key challenge in delivering services to this population is the high prevalence persons experiencing of homelessness which has not been previously demonstrated among PWID in India. PWID reporting homelessness, unstable housing, and migration may experience unmet needs for services and further disease transmission through socially and spatially dynamic networks.

A limitation of this cross-sectional analysis is that the reported networks and injection venues may not necessarily represent network members or venues where participants acquired HIV and/or HCV infection; however, this would likely bias observed associations toward the null and attenuated associations of sociometric and spatial factors. Further, the cross-sectional nature limits the ability to examine temporal associations, but the consistency of associations with active HCV infection suggests that these network factors may impact onward transmission. All responses related to drug use, network members, and spaces were self-reported and subject to social desirability and recall bias; to minimize bias, all interviewers were trained on optimal interviewing techniques. About 25% of referral coupons were not returned suggesting that the networks presented in these data may be incomplete; however, the response rate of 75% is higher than what has been seen in other network studies (Kimani et al., 2014; Johnston et al., 2008). PWID under the age of 18 were excluded from the study due to the legal age of consent in India; therefore, these individuals are not represented by the network topology. Statistical analyses assumed that observations are independent conditional on individual-level covariates. This assumption is likely to be violated to some extent, but violations are not expected to bias point estimates (they would result in underestimated standard errors).

Limitations notwithstanding, these data highlight the importance of networks on HIV and HCV burden in a community of PWID in New Delhi, India. Integrating strategies to intervene at sociometric- and spatial-levels in addition to individual-level interventions could improve the efficiency of prevention and treatment programming and may be critical to achieving epidemic control and elimination of HIV and HCV, respectively, while conserving resources.

An interactive version of the sociospatial network and underlying data are available from: https://github.com/sclipman/sociospatial-baseline.

1. Report

   1. Ambekar A
   2. Agrawal A
   3. Rao R
   4. Mishra A
   5. Khandelwal S
   6. Chadda R

   (2019)

   Magnitude of Substance Use in India

   Ministry of Social Justice and Empowerment Government of India.

   • Google Scholar
2. Software

   1. Clipman S

   (2021) sociospatial-baseline, version swh:1:rev:f22127448e931699530d02475043b2279279d67f

   Software Heritage.

   https://archive.softwareheritage.org/swh:1:dir:5d17c79c5c371aa593d45cc8bc7e9272b7b552ea;origin=https://github.com/sclipman/sociospatial-baseline;visit=swh:1:snp:281e1bee32d35ac218714d8b40ca179a94de0bd8;anchor=swh:1:rev:f22127448e931699530d02475043b2279279d67f
3. 1. Costenbader EC
   2. Astone NM
   3. Latkin CA

   (2006) The dynamics of injection drug users’ personal networks and HIV risk behaviors

   Addiction 101:1003–1013.

   https://doi.org/10.1111/j.1360-0443.2006.01431.x

   • Google Scholar
4. 1. DeHovitz J
   2. Uuskula A
   3. El-Bassel N

   (2014) The HIV epidemic in eastern Europe and central asia

   Current HIV/AIDS Reports 11:168–176.

   https://doi.org/10.1007/s11904-014-0202-3

   • PubMed
   • Google Scholar
5. 1. Des Jarlais DC
   2. Cooper HLF
   3. Arasteh K
   4. Feelemyer J
   5. McKnight C
   6. Ross Z

   (2018) Potential geographic "hotspots" for drug-injection related transmission of HIV and HCV and for initiation into injecting drug use in New York City, 2011-2015, with implications for the current opioid epidemic in the US

   PLOS ONE 13:e0194799.

   https://doi.org/10.1371/journal.pone.0194799

   • PubMed
   • Google Scholar
6. 1. Friedman SR
   2. Neaigus A
   3. Jose B
   4. Curtis R
   5. Goldstein M
   6. Ildefonso G
   7. Rothenberg RB
   8. Des Jarlais DC

   (1997) Sociometric risk networks and risk for HIV infection

   American Journal of Public Health 87:1289–1296.

   https://doi.org/10.2105/AJPH.87.8.1289

   • PubMed
   • Google Scholar
7. 1. Gesink D
   2. Wang S
   3. Norwood T
   4. Sullivan A
   5. Al-Bargash D
   6. Shahin R

   (2014) Spatial epidemiology of the syphilis epidemic in Toronto, Canada

   Sexually Transmitted Diseases 41:637–648.

   https://doi.org/10.1097/OLQ.0000000000000196

   • PubMed
   • Google Scholar
8. Report

   1. Hagberg A
   2. Swart P
   3. Chult D S

   (2008)

   Exploring Network Structure, Dynamics, and Function Using NetworkX

   Los Alamos, NM (United States: Los Alamos National Lab.(LANL).

   • Google Scholar
9. 1. Hellard M
   2. Rolls DA
   3. Sacks-Davis R
   4. Robins G
   5. Pattison P
   6. Higgs P
   7. Aitken C
   8. McBryde E

   (2014) The impact of injecting networks on hepatitis C transmission and treatment in people who inject drugs

   Hepatology 60:1861–1870.

   https://doi.org/10.1002/hep.27403

   • PubMed
   • Google Scholar
10. 1. Hellard M
    2. McBryde E
    3. Sacks Davis R
    4. Rolls DA
    5. Higgs P
    6. Aitken C
    7. Thompson A
    8. Doyle J
    9. Pattison P
    10. Robins G

    (2015) Hepatitis C transmission and treatment as prevention - The role of the injecting network

    International Journal of Drug Policy 26:958–962.

    https://doi.org/10.1016/j.drugpo.2015.05.006

    • PubMed
    • Google Scholar
11. 1. Johnston LG
    2. Khanam R
    3. Reza M
    4. Khan SI
    5. Banu S
    6. Alam MS
    7. Rahman M
    8. Azim T

    (2008) The effectiveness of respondent driven sampling for recruiting males who have sex with males in Dhaka, Bangladesh

    AIDS and Behavior 12:294–304.

    https://doi.org/10.1007/s10461-007-9300-1

    • PubMed
    • Google Scholar
12. 1. Kimani SM
    2. Watt MH
    3. Merli MG
    4. Skinner D
    5. Myers B
    6. Pieterse D
    7. MacFarlane JC
    8. Meade CS

    (2014) Respondent driven sampling is an effective method for engaging methamphetamine users in HIV prevention research in South Africa

    Drug and Alcohol Dependence 143:134–140.

    https://doi.org/10.1016/j.drugalcdep.2014.07.018

    • PubMed
    • Google Scholar
13. 1. Klovdahl AS
    2. Potterat JJ
    3. Woodhouse DE
    4. Muth JB
    5. Muth SQ
    6. Darrow WW

    (1994) Social networks and infectious disease: the Colorado springs study

    Social Science & Medicine 38:79–88.

    https://doi.org/10.1016/0277-9536(94)90302-6

    • PubMed
    • Google Scholar
14. 1. Kursa MB
    2. Rudnicki WR

    (2010) Feature selection with the boruta package

    Journal of Statistical Software 36:1–13.

    https://doi.org/10.18637/jss.v036.i11

    • Google Scholar
15. 1. Latkin C
    2. Donnell D
    3. Celentano DD
    4. Aramrattna A
    5. Liu TY
    6. Vongchak T
    7. Wiboonnatakul K
    8. Davis-Vogel A
    9. Metzger D

    (2009) Relationships between social norms, social network characteristics, and HIV risk behaviors in Thailand and the united states

    Health Psychology 28:323–329.

    https://doi.org/10.1037/a0014707

    • PubMed
    • Google Scholar
16. 1. Latkin CA
    2. Kuramoto SJ
    3. Davey-Rothwell MA
    4. Tobin KE

    (2010) Social norms, social networks, and HIV risk behavior among injection drug users

    AIDS and Behavior 14:1159–1168.

    https://doi.org/10.1007/s10461-009-9576-4

    • PubMed
    • Google Scholar
17. 1. Latkin C
    2. Yang C
    3. Srikrishnan AK
    4. Solomon S
    5. Mehta SH
    6. Celentano DD
    7. Kumar MS
    8. Knowlton A
    9. Solomon SS

    (2011) The relationship between social network factors, HIV, and hepatitis C among injection drug users in Chennai, India

    Drug and Alcohol Dependence 117:50–54.

    https://doi.org/10.1016/j.drugalcdep.2011.01.005

    • PubMed
    • Google Scholar
18. 1. Latkin CA
    2. Davey-Rothwell MA
    3. Knowlton AR
    4. Alexander KA
    5. Williams CT
    6. Boodram B

    (2013) Social network approaches to recruitment, HIV prevention, medical care, and medication adherence

    JAIDS Journal of Acquired Immune Deficiency Syndromes 63 Suppl 1:S54–S58.

    https://doi.org/10.1097/QAI.0b013e3182928e2a

    • PubMed
    • Google Scholar
19. 1. Logan JJ
    2. Jolly AM
    3. Blanford JI

    (2016) The sociospatial network: risk and the role of place in the transmission of infectious diseases

    PLOS ONE 11:e0146915.

    https://doi.org/10.1371/journal.pone.0146915

    • PubMed
    • Google Scholar
20. 1. Lucas GM
    2. Solomon SS
    3. Srikrishnan AK
    4. Agrawal A
    5. Iqbal S
    6. Laeyendecker O
    7. McFall AM
    8. Kumar MS
    9. Ogburn EL
    10. Celentano DD
    11. Solomon S
    12. Mehta SH

    (2015) High HIV burden among people who inject drugs in 15 indian cities

    Aids 29:619–628.

    https://doi.org/10.1097/QAD.0000000000000592

    • PubMed
    • Google Scholar
21. 1. Mehta SH
    2. Lucas GM
    3. Solomon S
    4. Srikrishnan AK
    5. McFall AM
    6. Dhingra N
    7. Nandagopal P
    8. Kumar MS
    9. Celentano DD
    10. Solomon SS

    (2015) HIV care continuum among men who have sex with men and persons who inject drugs in India: barriers to successful engagement

    Clinical Infectious Diseases : An Official Publication of the Infectious Diseases Society of America 61:1732–1741.

    https://doi.org/10.1093/cid/civ669

    • PubMed
    • Google Scholar
22. 1. Rolls DA
    2. Sacks-Davis R
    3. Jenkinson R
    4. McBryde E
    5. Pattison P
    6. Robins G
    7. Hellard M

    (2013) Hepatitis C transmission and treatment in contact networks of people who inject drugs

    PLOS ONE 8:e78286.

    https://doi.org/10.1371/journal.pone.0078286

    • PubMed
    • Google Scholar
23. 1. Rothenberg RB
    2. Woodhouse DE
    3. Potterat JJ
    4. Muth SQ
    5. Darrow WW
    6. Klovdahl AS

    (1995)

    Social networks in disease transmission: the Colorado springs study

    NIDA Research Monograph 151:3–19.

    • PubMed
    • Google Scholar
24. 1. Rudolph AE
    2. Young AM
    3. Havens JR

    (2017) Examining the Social Context of Injection Drug Use: Social Proximity to Persons Who Inject Drugs Versus Geographic Proximity to Persons Who Inject Drugs

    American Journal of Epidemiology 186:970–978.

    https://doi.org/10.1093/aje/kwx162

    • Google Scholar
25. 1. Solomon SS
    2. Mehta SH
    3. Srikrishnan AK
    4. Solomon S
    5. McFall AM
    6. Laeyendecker O
    7. Celentano DD
    8. Iqbal SH
    9. Anand S
    10. Vasudevan CK
    11. Saravanan S
    12. Lucas GM
    13. Kumar MS
    14. Sulkowski MS
    15. Quinn TC

    (2015) Burden of hepatitis C virus disease and access to hepatitis C virus services in people who inject drugs in India: a cross-sectional study

    The Lancet Infectious Diseases 15:36–45.

    https://doi.org/10.1016/S1473-3099(14)71045-X

    • PubMed
    • Google Scholar
26. 1. Solomon SS
    2. Boon D
    3. Saravanan S
    4. Srikrishnan AK
    5. Vasudevan CK
    6. Balakrishnan P
    7. Persaud D
    8. Ray SC
    9. Mehta S
    10. Mehta SH

    (2019a) Diversity of hepatitis C virus infection among HIV-infected people who inject drugs in India

    VirusDisease 30:490–497.

    https://doi.org/10.1007/s13337-019-00553-2

    • PubMed
    • Google Scholar
27. 1. Solomon SS
    2. Solomon S
    3. McFall AM
    4. Srikrishnan AK
    5. Anand S
    6. Verma V
    7. Vasudevan CK
    8. Balakrishnan P
    9. Ogburn EL
    10. Moulton LH
    11. Kumar MS
    12. Sachdeva KS
    13. Laeyendecker O
    14. Celentano DD
    15. Lucas GM
    16. Mehta SH
    17. Indian National Collaboration on AIDS Study

    (2019b) Integrated HIV testing, prevention, and treatment intervention for key populations in India: a cluster-randomised trial

    The Lancet HIV 6:e283–e296.

    https://doi.org/10.1016/S2352-3018(19)30034-7

    • PubMed
    • Google Scholar
28. 1. Young AM
    2. Jonas AB
    3. Mullins UL
    4. Halgin DS
    5. Havens JR

    (2013) Network structure and the risk for HIV transmission among rural drug users

    AIDS and Behavior 17:2341–2351.

    https://doi.org/10.1007/s10461-012-0371-2

    • PubMed
    • Google Scholar
29. 1. Zelenev A
    2. Long E
    3. Bazazi AR
    4. Kamarulzaman A
    5. Altice FL

    (2016) The complex interplay of social networks, geography and HIV risk among malaysian drug injectors: results from respondent-driven sampling

    International Journal of Drug Policy 37:98–106.

    https://doi.org/10.1016/j.drugpo.2016.08.008

    • PubMed
    • Google Scholar
30. 1. Zelenev A
    2. Li J
    3. Mazhnaya A
    4. Basu S
    5. Altice FL

    (2018) Hepatitis C virus treatment as prevention in an extended network of people who inject drugs in the USA: a modelling study

    The Lancet Infectious Diseases 18:215–224.

    https://doi.org/10.1016/S1473-3099(17)30676-X

    • PubMed
    • Google Scholar

Acceptance summary:

This study explores the impact of social networks of injecting drug users on the risk of hepatitis C virus and HIV transmission. The study concludes that social and spatial networks of injecting drug users have a significant influence on their risk of becoming hepatitis C virus or HIV positive, on top of their individual risk factors. This paper is of interest for epidemiologists and those who are working for people who inject drugs and people living with HIV.

Decision letter after peer review:

Thank you for submitting your article "Role of direct and indirect social and spatial ties in the diffusion of HIV and HCV among people who inject drugs: A cross-sectional community-based network analysis in New Delhi, India" for consideration by eLife. Your article has been reviewed by 2 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Miles Davenport as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Chaturaka Rodrigo (Reviewer #2).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

1. In egocentric network, the connection between first degree alters is referred by an index, but such information may not accurate (alters may or may not connected regardless of the information from the index). Were these connections confirmed by the first degree alters? This is not just the issue in the information provided by indexes. It is generally challenging to confirm connections. Some may not share their names nor background because of various reasons beyond those described by the authors: 1. they do not need such information to share syringes, and 2. they are afraid that the information is transferred to law enforcement agencies, or 3. simply there are too many to refer (I do not think it is realistic to ask each person whether all 2511 participants are connected to him/her). The diameter and average path length were dramatically reduced by incorporating injection location, which might be explained by the incompleteness of the network, because undocumented connections could be explained by locations as the authors described.

2. The authors used "number infected injection partners" and "network distance from an HIV/HCV RNA⁺ Person" as predictor of HIV/HCV infection. It was not clear why these two factors which summarize the network structure were selected, although they sound reasonably associated with infection risk.

3. Is it possible to use different shapes to distinguish the participants and the hotspots?

4. Table 1 – Please explain the meaning of "anti-HCV" as a footnote in the table.

5. Table 1 – What is the relevance of "Ever used non-injecting drugs"? If this information needs to be put here, it is much better to state non-jecting drug use in the preceding six months.

6. "total 155 (86%) of the 181 injection locations included at least one HIV positive person" – This sentence is unclear.

7. "Individual-level variables associated with prevalent HIV in multivariable logistic regression included age, education, homelessness, sexual activity, needle sharing, injection frequency, and type of drug injected" – It is better to indicate the direction of association in this sentence (e.g. younger age is associated with HIV). While the details can be worked out from table 2, it is easier for the reader if this is mentioned here.

8. If a person had anti-HCV antibodies but no HCV RNA, did you ask about treatment?

9. "persons with active HCV infection had significantly higher degree centrality" – Is this for all edges (HCV and HIV) or edges from HCV infected nodes only?

10. Lines 175 – 181: Please provide adjusted OR with 95% CI for each % value as you have done for HIV section.

11. Line 241 – What is "Deep chain RDS"?

12. People under 18 has been excluded due to legal age of consent- This can have a significant effect on the network topology – please acknowledge this as a limitation.

13. Please provide sensitivity, specificity, NPV (negative predictive value) and PPV of tests you have used to diagnose HCV and HIV.

14. It is next to impossible to understand what you meant in the Results section by "Median network distance to another HIV antibody-positive PWID was 0", without the following explanation that appears in methods – "Network distance from an infected alter was calculated such that a distance of zero signifies a direct connection, a distance of one signifies one uninfected person along the shortest path between a given participant and infected alter" The latter information should also appear in results where you discuss median network distance.

15. Will there be any legal implications in publishing the map in Supplementary figure 3 – It may reveal the location of injecting hotspots to law enforcement, drive injection networks underground, and hinder future work of this cohort study. It will also have a negative impact on HCV and HIV prevention strategies on these locations.

16. Can another dimension be incorporated into the networks to indicate the impact of preventive measures over time? – This is a suggestion for future work and not necessarily to be addressed in this paper.Reviewer #1 (Recommendations for the authors):

The authors investigated the association between HIV/HCV infection risk and network structure (closeness to those infected). To rigorously investigate the network structure, they employed biometric data (fingerprint) along with identifying information of referred persons. They further coupled the network data with the information of where the participants inject drugs, which is to complement the network structure as well as to investigate the unique contribution of the special information to the infection risk. They found both network structure and spatial information explain infection risk even after controlling for individual-level risk factors. There are some studies investigated the network structure and its contribution to the infection risk of HIV/HCV, however, this study is unique as they used rigorous approach to reconstruct the network structure beyond egocentric network with substantial effort. This study gives insights to the network-based intervention to control HIV/HCV epidemics implemented around the world.

The strength of this study is that they recruited large number of participants and the effort to complete the network: they collected the information of network carefully avoiding duplicates using not just identifying information but also biometric data (fingerprint). Meanwhile, the limitation of this study is, as is described by the authors, incompleteness of the network.Reviewer #2 (Recommendations for the authors):

This study was well designed to achieve its objective of uncovering a broader social and spatial network of injecting drug users in New Delhi, India. The extent of the network uncovered is impressive starting with just 10 index cases. The reliability of data is high given that 75% of referrals distributed through this network was returned to investigators. The authors have maintained rigorous standards in methodology as far as practically possible for a study of this scale conducted in one of the most densely populated places on earth.

The methods explained here in my opinion, will set a new standard in research that explores a person's risk of blood borne viral infections (HCV/HBV and HIV). The results show that in addition to typically reported individual risk factors such as age and education etc, the social and spatial network of an injecting drug user also significantly influences their risk of HIV/HCV infection. This finding will be a boost to community-based prevention methods by providing evidence to back their existence and relevance.

The study has some limitations – findings are not generalisable outside the geography studied but these methods can be employed to study other areas of interest. The study has excluded persons under the age of 18, which can have a significant impact on the topology of the network in the form of missing links.

Essential revisions:

1. In egocentric network, the connection between first degree alters is referred by an index, but such information may not accurate (alters may or may not connected regardless of the information from the index). Were these connections confirmed by the first degree alters? This is not just the issue in the information provided by indexes. It is generally challenging to confirm connections. Some may not share their names nor background because of various reasons beyond those described by the authors: 1. they do not need such information to share syringes, and 2. they are afraid that the information is transferred to law enforcement agencies, or 3. simply there are too many to refer (I do not think it is realistic to ask each person whether all 2511 participants are connected to him/her). The diameter and average path length were dramatically reduced by incorporating injection location, which might be explained by the incompleteness of the network, because undocumented connections could be explained by locations as the authors described.

We apologize for any confusion, however, the connections between first degree alters is not referred by an index as the reviewer suggests i.e., an index did not determine if two of their partners should be connected to each other. In this study network connections were only established by return of a referral card. Sampling was initiated with 10 indexes. The indexes were asked to name individuals who they injected with in the prior month (regardless of whether they shared injection paraphernalia) as well as provide a factoid about each individual (e.g., “he is originally from Bihar” or “she has a scar under the left eye”). They were then provided unique barcoded referral cards for each of the injection partners they named (there was no limit on the number of partners one could name), and they were asked to provide the card(s) to their partner(s). When an individual returned a referral card, a biometric fingerprint scan was used to determine if they had been previously enrolled in the study. If previously enrolled, the referral card was used to establish an additional cross-network link between the individual and person who referred them. If the individual was not previously enrolled, the factoid previously provided was used to confirm they were indeed the partner described and the referral card was used to establish a network connection with the participant that referred them. This recruited participant was next considered an “index” and in turn asked to name and provide a factoid about the persons they injected with in the prior month and was provided referral cards for each of these partners. They were then asked to distribute the cards to their partners and when the cards were returned, network connections were established and the process was repeated. Therefore, for a network connection (edge) to be created between two participants (nodes) there had to be a direct barcoded referral card linking the two participants; connections were not inferred by asking an index if his/her alters injected with each other. Moreover, cross-network linkages between individuals referred by different indexes were established via biometric linkage. Although 75% of all referral cards distributed were returned, we agree with the reviewer that additional undocumented connections may exist for a multitude of reasons. The key reasons raised by the reviewer include:

1) they do not need such information to share syringes and (2) they are afraid that the information is transferred to law enforcement agencies, or (3) simply there are too many to refer (I do not think it is realistic to ask each person whether all 2511 participants are connected to him/her).

Participants were asked to provide referral cards to individuals regardless of whether they share syringes and did not need to provide real names in the case they did not know the person well or were afraid that information would be transferred to law enforcement agencies. Our research group has spent over 10 years developing trust and rapport with this PWID community. We conducted thorough ethnographic research and met with community leaders and PWID across the city, and this was not a concern expressed by participants. Individuals may not recall all their injection partners; however, we asked several additional questions to get estimates of the number of PWID individuals know and have injected with and trained interviewers to clarify and revisit questions in the case of inconsistencies. Moreover, prior work across India has supported that the number of injection partners within the prior month is relatively small- a median of 2 partners. This is fairly consistent across cities where we have worked in India. Moreover, we were asking participants to recall people they specifically injected with not just had passing contact with. We used the past month to minimize these recall issues.

Additionally, 75% network completeness is high for network studies among PWID, and we believe that the addition of spaces individuals inject allow us to further model undocumented connections. The network diameter and average path length were reduced by incorporating injection location, demonstrating the utility of spaces in reaching network participants, and we agree with the reviewer that this might be explained by the incompleteness of the network, but it also does not strictly signify incompleteness. Given a toy example where all injection partner connections are known, creating a bi-partite network with spaces can still result in reduced pathlength and diameter, irrespective of completeness. For example, in Author response image, 1 of a hypothetical example of a complete network, the diameter of (a) is 3 since the longest shortest paths have length of 3 edges (e.g., between 6-1 and 6-2); however, if we create a bi-partite network with a common space depicted as a blue square in (b), the diameter is reduced to 2.

Author response image 1

Download asset Open asset

2. The authors used "number infected injection partners" and "network distance from an HIV/HCV RNA⁺ Person" as predictor of HIV/HCV infection. It was not clear why these two factors which summarize the network structure were selected, although they sound reasonably associated with infection risk.

We considered candidate risk factors based on biological/epidemiological significance and prior literature and also used a Boruta random forest feature selection algorithm to comprehensively explore candidate factors. Both the number of infected partners and network distance to an infected partner had significant variable importance scores from random forest (p<0.05) and are epidemiologically meaningful since an increase in these variables represents an increased number of pathways through which transmission could occur, as demonstrated in several treatment as prevention studies for HIV. While existing studies have reported associations with HIV/HCV infection and the number of infected partners, network distance to an infected partner may help to explain the increased risk of transmission even before a direct connection (injection partner) is diagnosed or represent viremia in the extended partner network, further approximating the chance that an unreported or future injection partner is infected, which to our knowledge is novel.

3. Is it possible to use different shapes to distinguish the participants and the hotspots?

We thank the reviewer for this excellent suggestion. The figure has been revised to display injection locations as squares.

4. Table 1 – Please explain the meaning of "anti-HCV" as a footnote in the table.

We have added the meaning of anti-HCV as a footnote in the table.

5. Table 1 – What is the relevance of "Ever used non-injecting drugs"? If this information needs to be put here, it is much better to state non-jecting drug use in the preceding six months.

Thank you, we have removed "Ever used non-injecting drugs" from table 1.

6. "total 155 (86%) of the 181 injection locations included at least one HIV positive person" – This sentence is unclear.

We have revised this sentence in the manuscript to read:

“At least one HIV-positive person reported injecting at 155 (86%) of the 181 injection locations identified by participants.”

7. "Individual-level variables associated with prevalent HIV in multivariable logistic regression included age, education, homelessness, sexual activity, needle sharing, injection frequency, and type of drug injected" – It is better to indicate the direction of association in this sentence (e.g. younger age is associated with HIV). While the details can be worked out from table 2, it is easier for the reader if this is mentioned here.

Thank you, we appreciate this suggestion and have revised the manuscript to indicate the direction of association in this sentence. The revised sentence reads:

“Individual-level variables positively associated with prevalent HIV in multivariable logistic regression included younger age, lower education, experiencing homelessness, decreased sexual activity, sharing syringes, increased injection frequency, and injecting heroin and buprenorphine (Table 2).”

8. If a person had anti-HCV antibodies but no HCV RNA, did you ask about treatment?

We asked all participants about prior HCV testing and treatment. The majority of these PWID were unaware of their HCV status, only 104 (4%) had ever been previously tested for HCV; 9 individuals reported ever testing positive (all of whom received treatment or were currently taking treatment), and 3 reported a sustained virologic response. Therefore, instances where a person had anti-HCV antibodies but no HCV RNA most likely represent natural/spontaneous clearance of HCV infection. We have revised the manuscript to include this information.

9. "persons with active HCV infection had significantly higher degree centrality" – Is this for all edges (HCV and HIV) or edges from HCV infected nodes only?

We used the standard definition of degree centrality i.e., the total number of edges a node has, irrespective of HIV and/or HCV status.

10. Lines 175 – 181: Please provide adjusted OR with 95% CI for each % value as you have done for HIV section.

We thank the reviewer for catching this and have revised the manuscript to include the adjusted OR with 95% CI for each value.

11. Line 241 – What is "Deep chain RDS"?

We thank the reviewer for pointing out that the acronym for respondent-driven sampling (RDS) was not previously defined. We have revised the manuscript to define RDS and have also rearranged the citations in the sentence to better detail which reference corresponds to “deep chain RDS” should the reader be interested in learning more. “Deep chain RDS” is an approach that hypothesizes that the deeper the recruitment depth in a RDS sample, the more likely one is to identify HIV+ persons not engaged in care.

12. People under 18 has been excluded due to legal age of consent- This can have a significant effect on the network topology – please acknowledge this as a limitation.

We appreciate the reviewer highlighting this and have revised the manuscript to acknowledge this as a limitation.

13. Please provide sensitivity, specificity, NPV (negative predictive value) and PPV of tests you have used to diagnose HCV and HIV.

We have added details on the sensitivity and specificity for each test used to diagnose HCV and HIV to the Materials and methods section.

14. It is next to impossible to understand what you meant in the Results section by "Median network distance to another HIV antibody-positive PWID was 0", without the following explanation that appears in methods – "Network distance from an infected alter was calculated such that a distance of zero signifies a direct connection, a distance of one signifies one uninfected person along the shortest path between a given participant and infected alter" The latter information should also appear in results where you discuss median network distance.

We thank the reviewer for pointing this out. The article structure has been reordered so that methods now come before the results, which should remedy this.

15. Will there be any legal implications in publishing the map in Supplementary figure 3 – It may reveal the location of injecting hotspots to law enforcement, drive injection networks underground, and hinder future work of this cohort study. It will also have a negative impact on HCV and HIV prevention strategies on these locations.

We understand and acknowledge the reviewer’s concern and have further anonymized this figure (Figure 3—figure supplement 1). We have replaced the map with a solid white background so injection locations cannot be geographically placed. The distance between points is still geospatially congruent to give the reader an idea of their distribution and HIV/HCV prevalence.

16. Can another dimension be incorporated into the networks to indicate the impact of preventive measures over time? – This is a suggestion for future work and not necessarily to be addressed in this paper

Thank you for this excellent suggestion, we are currently in the process of modeling the impact of various preventive measures over time and plan to publish an additional paper on these findings.

Author details

1. Steven J Clipman

   Department of Medicine, Division of Infectious Diseases, Johns Hopkins School of Medicine, Baltimore, United States

   Contribution

   Conceptualization, Data curation, Software, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing - original draft, Writing - review and editing

   For correspondence

   sclipman@jhmi.edu

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-2366-8420

2. Shruti H Mehta

   Department of Epidemiology, Johns Hopkins Bloomberg School of Public Health, Baltimore, United States

   Contribution

   Conceptualization, Supervision, Funding acquisition, Methodology, Writing - original draft, Writing - review and editing

   Competing interests

   reports personal fees from Gilead Sciences, outside the submitted work.

3. Aylur K Srikrishnan

   YR Gaitonde Centre for AIDS Research and Education (YRGCARE), Chennai, India

   Contribution

   Conceptualization, Supervision, Funding acquisition, Investigation, Project administration, Writing - review and editing

   Competing interests

   No competing interests declared

4. Katie JC Zook

   Department of Medicine, Division of Infectious Diseases, Johns Hopkins School of Medicine, Baltimore, United States

   Contribution

   Data curation, Software, Validation, Investigation, Project administration, Writing - review and editing

   Competing interests

   No competing interests declared

5. Priya Duggal

   Department of Epidemiology, Johns Hopkins Bloomberg School of Public Health, Baltimore, United States

   Contribution

   Supervision, Writing - review and editing

   Competing interests

   No competing interests declared

6. Shobha Mohapatra

   YR Gaitonde Centre for AIDS Research and Education (YRGCARE), Chennai, India

   Contribution

   Data curation, Supervision, Investigation, Project administration, Writing - review and editing

   Competing interests

   No competing interests declared

7. Saravanan Shanmugam

   YR Gaitonde Centre for AIDS Research and Education (YRGCARE), Chennai, India

   Contribution

   Investigation, Project administration, Writing - review and editing

   Competing interests

   No competing interests declared

8. Paneerselvam Nandagopal

   YR Gaitonde Centre for AIDS Research and Education (YRGCARE), Chennai, India

   Contribution

   Investigation, Project administration, Writing - review and editing

   Competing interests

   No competing interests declared

9. Muniratnam S Kumar

   YR Gaitonde Centre for AIDS Research and Education (YRGCARE), Chennai, India

   Contribution

   Supervision, Funding acquisition, Investigation, Project administration, Writing - review and editing

   Competing interests

   No competing interests declared

10. Elizabeth Ogburn

    Department of Biostatistics, Johns Hopkins Bloomberg School of Public Health, Baltimore, United States

    Contribution

    Conceptualization, Formal analysis, Supervision, Validation, Methodology, Writing - review and editing

    Competing interests

    No competing interests declared

11. Gregory M Lucas

    Department of Medicine, Division of Infectious Diseases, Johns Hopkins School of Medicine, Baltimore, United States

    Contribution

    Conceptualization, Supervision, Funding acquisition, Methodology, Project administration, Writing - review and editing

    Competing interests

    No competing interests declared

12. Carl A Latkin

    Department of Health, Behavior, and Society, Johns Hopkins Bloomberg School of Public Health, Baltimore, United States

    Contribution

    Conceptualization, Supervision, Methodology, Writing - review and editing

    Competing interests

    No competing interests declared

13. Sunil S Solomon

    1. Department of Medicine, Division of Infectious Diseases, Johns Hopkins School of Medicine, Baltimore, United States
    2. Department of Epidemiology, Johns Hopkins Bloomberg School of Public Health, Baltimore, United States

    Contribution

    Conceptualization, Resources, Data curation, Supervision, Funding acquisition, Validation, Investigation, Methodology, Writing - original draft, Project administration, Writing - review and editing

    For correspondence

    sss@jhmi.edu

    Competing interests

    reports grants/products and advisory board fees from Gilead Sciences and grants/products from Abbott Diagnostics, outside the submitted work.

• 843

  Page views

• 93

  Downloads

• 3

  Citations

Article citation count generated by polling the highest count across the following sources: PubMed Central, Crossref, Scopus.