There are an estimated 75,000 to 125,000 injection drug users (IDUs) in Canada.1 In the city of Montreal, as in many other Canadian cities, the most commonly injected drugs are typically cocaine and heroin. Regional surveillance data indicate that 70% of the estimated 13,600 IDUs in Montreal inject cocaine as their primary drug and 23% inject heroin, whereas 7% inject other drugs.2
The choice of drug is often mediated by factors such as drug availability, drug culture, regional drug use habits, and individual drug user preferences.3,4 It is known that the drug of choice influences the frequency and type of interactions between injecting partners, which may partly account for the differential rates of HIV, hepatitis C virus (HCV), and other bloodborne infections observed in various IDU populations.5-7 Although cocaine and heroin injection have both been associated with HIV infection, some evidence suggests that cocaine injectors have risk factors that make them more prone to HIV and HCV infection than users of other drugs.8-14 Several markers of social disadvantage, such as low income, low educational attainment, homelessness, and unemployment, are linked to cocaine use.15 The use of shooting galleries, frequent needle sharing, having a high number of needle-sharing partners, and less consistent bleaching of injecting equipment are also well documented among cocaine injectors.8,9,15-21
The risk behaviors associated with cocaine use have been linked to the pharmacologic properties of the drug. Cocaine produces a rapid short-acting euphoria followed by reduced inhibition and judgment. Although cocaine and heroin both act on the dopaminergic pathways in the brain, they have different physiologic effects. Use of a particular drug may correspond to different patterns of consumption, physiologic dependence, risk behaviors, and risk reduction.22 IDUs who inject cocaine or speedball (a mixture of cocaine and heroin) often inject more frequently than users of heroin, because the effects of the former are short-lived. Frequent and erratic cocaine injections increase the opportunity for acquiring and transmitting bloodborne infections.
For many IDUs, cocaine use is characterized by binging, which profoundly influences an IDU's risk of bloodborne infections.13,23-25 The psychologic and physical condition of the individual during cocaine binges is significantly impaired because of the compulsion to obtain more drugs, resulting in long periods of sleep deprivation, inadequate nutrition, and depression.26 In a study of IDUs in Vancouver, frequent cocaine injection, compared with heroin injection, was associated with an approximate 4-fold higher risk for HIV infection during a 1-year follow-up.10 Furthermore, the study found a dose-response relation between the risk of HIV seroconversion and intensity of cocaine injection.
Although prior studies suggest that bloodborne infection risk is a function of drug type, no previous research has examined how infection risk can be differentiated by the social networks that are defined by drug type. There is increasing evidence that an IDU's social network and, more importantly, the drug-injecting network is important in conceptualizing risk for individual IDUs, because network members can influence drug use behaviors of their peers. For example, injecting network characteristics, such as the size of one's drug-injecting network, the frequency of contact with other IDUs, and the change in the injecting network membership, have been associated with HIV risk.27-29
Therefore, the objective of our study was to examine differences in risk factors for bloodborne infections by comparing the social and, in particular, the drug-injecting networks of cocaine and heroin injectors.
Study Design and Population
Active IDUs who injected at least once in the past 6 months were recruited into this cross-sectional study from 3 of the largest syringe exchange programs (SEPs) and from 2 methadone maintenance treatment (MMT) clinics in Montreal, Canada between April 2004 and January 2005. Methadone clients in this sample were newly initiated into treatment (median time in treatment = 2 months). As such, it was expected that they would be similar to IDUs not currently in treatment with regard to their injection risk behaviors,30 because many such IDUs continue to inject during the early months of treatment and maintain pretreatment risk behaviors.31 Indeed, the sample of heroin injectors from MMT clinics did not differ significantly from the nontreatment sample with regard to the drug use variables examined in this study.
The study was part of a larger project aimed at understanding hepatitis C-related psychosocial factors and risk behaviors of IDUs in Montreal. The project was promoted by posting flyers on bulletin boards at recruitment sites, word of mouth, and site personnel. Systematic sampling of every second client seeking services at the recruitment sites was used to minimize selection bias. A uniform recruitment strategy was used at all participating sites. Interviews were conducted in a private room on site or arranged at the Montreal Regional Public Health Department whenever appropriate.
Eligibility as an active IDU was verified by presence of injection marks or through knowledge about community services offered to IDUs and of typical injection procedures. Participants were at least 18 years of age, provided informed consent, and were reimbursed CDN $20 for their time. Subjects underwent an anonymous face-to-face interview that lasted, on average, 1.5 hours. Data were collected using a structured questionnaire administered by trained study interviewers who were not affiliated with the recruitment sites. At the conclusion of the interview, participants were given general information about HCV infection. All persons approached for participation were offered referrals to community services for IDUs.
The study procedures were approved by the McGill University Faculty of Medicine Institutional Review Board for Research on Human Subjects.
Participants' sociodemographics (age, gender, education, income, and housing), drug preparation and injection practices (number of drugs injected, age at first injection, years injecting, place and frequency of injecting, and sharing of syringes), and self-reported HIV and HCV infection status were elicited. Public or semipublic places of injecting were defined as areas without privacy, such as in the street, cars, parks, public toilets, and abandoned buildings. Private injection settings referred to one's own home, the home of a friend or family member, or a hotel/motel room. Sharing was defined as borrowing or lending injecting equipment previously used by another injector. All risk behavior variables referred to the past 6 months or 1 month before study enrollment. The 1-month time frame of risk behavior assessment was used for questions most likely to be affected by recall.
Subjects were deemed to be primarily cocaine or heroin injectors if either of these drugs was injected “half the time or more” during the past 6 months. This dichotomy of injection frequency was chosen because we wished to distinguish between infrequent users who reported using a specific drug, such as cocaine, sometimes or rarely from regular users of the drug who used half the time (6% of all cocaine injectors), most of the time (25%), or always (44%).
Elicitation of the Social Network
The network questionnaire was modeled on the General Social Survey and the National Institute on Drug Abuse (NIDA) Risk Behavior Assessment Questionnaire, which were designed to capture the effect of social networks on the behaviors of study participants and the risk practices of IDUs.32,33 Each study participant (also called an index) was asked to identify up to 10 individuals anonymously (IDUs or non-IDUs) with whom they had significant interaction during the past month. Significant interaction was defined as having more than casual contact, in which the nominated network member played an important role in the subject's life. It is worthy to note that the number of additional network members beyond the initial 10 was also recorded, but no information about these members was collected. Interviewers were trained to use memory-enhancing cues to elicit the nomination of members and to provide examples of roles of persons who might represent network members.34,35 Participants were asked to classify each of their network members into 1 or more relationship domains: IDU, sexual partner (regular or casual), sexual client, family member, other social support (eg, friend, coworker), drug dealer, and/or acquaintance. Network members could be classified in multiple categories. Each of these domains was reclassified as support (consisting of family and other social support), IDUs (any person reported as using drugs together with the index), or sexual partners (casual and regular, not including sexual clients).
Once the list of network members was generated, participants were asked to characterize each member by age, gender, ethnicity, duration of relationship, HIV and HCV status, and history of injection drug use. For a network member identified as a current IDU, questions were asked about the member's duration of injecting, type of drugs currently used, place and frequency of injection, history of shooting gallery attendance, current use of SEPs or drug treatment services, and sharing of injecting equipment with the index. Variables describing the structure of the IDU network included the size of the network (derived from the number of injecting individuals in the network) and network turnover (operationalized as the proportion of new IDUs in a network in the past month relative to the IDUs in the network during the past 6 months).
Because few participants reported using speedball (n = 14), we excluded them from analyses as they were not significantly different in their sociodemographic profile from the subjects included in the study. In addition, since we were interested in examining social network characteristics, we excluded subjects who did not report any social network members.
Descriptive statistics were calculated, with continuous variables compared using the Student t test and categoric variables examined with the Pearson χ2 test. Variables were considered to be statistically significant at P < 0.05 (2-tailed).
Variables to be modeled were selected based on prior substantive knowledge presented in the medical literature and on hypothesized associations. Bivariate and multivariate regression analyses were performed using generalized estimating equations (GEE) to account for clustering of network members on the index.36 Unadjusted and adjusted odds ratios (ORs) (with 95% confidence intervals [CIs]) were calculated for the association between cocaine injection (using heroin as the referent group) and personal and network characteristics. Independent variables found to be significant (P < 0.20) in the bivariate analysis were examined in a multivariate model. Interactions between significant covariates and age and gender were considered. Two models were examined: the first comparing the social networks and the second comparing the drug-injecting networks of cocaine and heroin injectors. The final models were chosen based on the quasi-likelihood statistic after sequentially removing variables from the saturated model and retaining significant variables (P < 0.05).
Of the 282 participants, 228 (81%) injected cocaine and 54 (19%) injected heroin as their primary drug in the past 6 months. Most index subjects (83%) were recruited from SEPs. The mean age of the study sample was 33 years; participants were predominantly male (73%), white (90%), and single (88%); and many lived in unstable housing conditions (42%). Based on self-report, 19% were HIV-positive only, 64% were HCV-positive only, and 19% were coinfected with HIV and HCV. As shown in Table 1, cocaine injectors were more likely than heroin users to demonstrate several markers of social disadvantage, including lower education, unstable housing, and receiving social assistance. Cocaine users were also significantly more likely to be male, to be HIV- or HCV-positive or coinfected, and to share syringes. Conversely, they were less likely to report injecting daily and polydrug use compared with heroin injectors.
Social and Drug-Injecting Networks
Study participants identified 714 social network members, of whom 364 were IDUs (305 cocaine users and 59 heroin users), 383 were providers of social support, and 169 were sexual partners. Overall, the mean social network sizes of cocaine injectors and heroin injectors were 4.22 (range: 1 to 10) members and 3.30 (range: 1 to 8) members, respectively. Among cocaine injectors, 9 index subjects reported more than 10 (median = 10, range: 2 to 30) network members. There was a high degree of concordance in the drug of choice between subjects and their drug-injecting network members (Table 2), supporting the notion of drug use homogeneity in injecting networks. The distribution of network member roles was found to differ significantly by drug type, because a greater proportion of support members were observed in the social networks of heroin injectors while more IDVs comprised the networks of cocaine injectors. The networks of cocaine injectors were also distinctly different with regard to the gender, age, HIV and HCV infection status, network size, duration of relationship and injection practices of their IDU network members.
Personal and Social Network Factors Associated With Cocaine Injection
Bivariate regression analysis identified several social network variables that were further considered in multivariate models. Table 3 shows results of the multivariate GEE regression analyses of social network factors associated with cocaine injectors. The social network model demonstrated that compared with heroin use, cocaine injection was associated with unstable housing (OR = 3.55, 95% CI: 1.49 to 8.40), less polydrug use (OR = 0.06, 95% CI: 0.02 to 0.16), HCV-positive status (OR = 4.69, 95% CI: 2.14 to 10.31), fewer social support members (OR = 0.97, 95% CI: 0.95 to 0.99 per member), and greater IDU network size (OR = 1.61, 95% CI: 1.14 to 2.28 per member).
A second model that considered only the drug-injecting networks of cocaine versus heroin injectors (Table 4) revealed that older age of network members (OR = 1.08, 95% CI: 1.04 to 1.12 per year), a history of shooting gallery attendance by members (OR = 2.27, 95% CI: 1.08 to 4.76), and shorter relationships of members with the index (OR = 0.91, 95% CI: 0.85 to 0.97) were associated with cocaine injection.
One of the strengths of a network approach to studying risk factors is the ability to examine the extent to which individuals align themselves into distinct networks. The degree of segregation and homogeneity of these networks can determine the ease with which bloodborne viruses are spread in a population. In this study, we examined how affiliation with cocaine or heroin networks was associated with HIV and HCV infection risk. Our results confirm findings from previous studies showing that cocaine injectors have several personal traits associated with a higher probability for HIV infection.8,9 Markers of marginalization, such as living in unstable housing, have been described in other predominantly cocaine-injecting populations.10 Lower access to resources for acquiring drugs may explain why fewer cocaine users in this sample were polydrug users. The higher self-reported prevalence of HCV among cocaine injectors is likely also indicative of greater injection risk behaviors in this group. Among IDUs in Sydney, Australia, higher HCV prevalence among cocaine and speedball users compared with heroin users suggested that injection practices and interaction patterns between injectors in various IDU networks can create environments that offer varying reproductive potentials for different pathogens.37
In addition to differing personal characteristics, we found that cocaine injectors were less likely to have social support and more likely to have larger IDU networks. Injection risk behaviors are likely to be sustained when there is less exposure to social support members who can encourage reduction of drug use as opposed to pressure from drug-using peers who encourage continued injecting.28 Indeed, belonging to a larger IDU network is associated with greater risk of HIV and HCV.27,28 Larger IDU networks may provide greater peer pressure for needle sharing and more sources for used injecting equipment. In the current study, the size of the social support network seemed to be less important than the mere presence of social support. The reverse was true for IDU networks, for which the size of the injecting network was more important than the presence of an injector in one's social network. It is possible that heroin and cocaine injectors differed in the amount of interaction with non-drug-using individuals, even though both groups had some injectors in their social networks.
The characteristics of drug network members also provide evidence for higher risk of HIV and HCV. Although the turnover rate suggested no difference in network membership change, the duration of relationships showed that cocaine injectors knew their injecting partners for a shorter period of time compared with heroin users. We suspect that the measure of change using the turnover rate (measured for the past 6 months) was too short to capture the dynamics of a network compared with the duration of relationships, which measured social network change over a lifetime period. Regardless, unstable relationships in drug-injecting networks have previously been associated with a higher probability of HIV transmission.29,38 This is not an unlikely finding, because evidence from epidemiologic and ethnographic research indicates that social relationships of IDUs are generally transient and short term.29,38-41 In one study, for example, the names of drug-using partners changed significantly more over time than the names of sexual partners or friends.38
The members of cocaine networks were also more likely to have a history of shooting gallery use, which has been associated with a high prevalence of risk behaviors for HIV and HCV infection.42 Such locations where IDUs congregate to inject have been linked to the selling, renting, and sharing of injection equipment while also exposing participants to networks of anonymous IDUs.43,44 The positive association with shooting galleries demonstrates that IDUs may seldom realize the full extent of their risk for infection arising from activities of their network members.45,46 Moreover, they would be unaware of the behaviors of other IDUs with whom their network members may associate but with whom they themselves have no direct contact.
The injection behaviors of opiate users have been found to be more consistent and involve less risk.47,48 The daily use of heroin and the need to use less frequently lend themselves to more predictable sources of drugs, and perhaps more organization regarding the acquisition and orderly preparation of drugs compared with the use of cocaine.10 However, as suggested by studies on IDUs who use multiple injectable drugs, there is likely a certain degree of mixing between members of different drug-using networks.47 Social assimilation of drug injectors into high-risk subpopulations may create conditions that help to perpetuate disease transmission. This bridging of members between networks of higher and lower risk therefore has implications for the spread of disease.44 An interesting topic for future research would be to examine how bridging individuals differ with regard to personal and network risk factors, given their unique network position.
The results of this study must be considered in light of several limitations. Because of the cross-sectional nature of the study, the observed social network structure might not be representative of past networks. Second, drug use patterns of cocaine and heroin injectors may be more complex than described, because some injectors are likely to use multiple drugs.47 Third, past network studies have found varying validity and reliability of self-reported information on characteristics of self and network members.49-52 Moreover, the extent to which participants accurately name all the persons in their network is a potential limitation of all network studies.53,54 IDUs tend to list earlier in recall the injecting partners with whom more frequent contact occurs.35 Although weak ties can be important for influencing behavior,45 it has been documented that a significant number of people's weaker social ties are often not remembered or not reported in standard network surveys.51,55 Finally, this sample of IDUs recruited predominantly from SEPs may not be representative of the IDU population at large. The monetary incentive offered for participation may have attracted IDUs who are in more financially precarious situations and whose behaviors are not representative of the injecting population at large. Unmeasured differences between cocaine and heroin injectors that could also bias the results include the frequency of contact with preventive services, the extent of behavioral interaction with other drug networks, and differences in awareness of HIV or HCV risk behaviors.
The results of this study stress the importance of considering risk influences beyond those of individual IDUs. Public health disease control measures must also recognize that distinct social networks may exist based on cocaine, heroin, and other drug types and that different affiliations can determine one's risk for bloodborne virus infection. Given the instability of cocaine injectors' relationships found in this study, network-based dissemination of information or sterile injecting materials may have limited impact. For example, in Vancouver and Montreal, the exchange of sterile injecting equipment between IDUs was found to be more common among heroin injectors compared with cocaine users.10 The authors of the study hypothesized that IDUs who primarily injected heroin were more consistent with their pattern of drug use and more easily formed networks through which harm reduction behaviors could be transmitted. Given their propensity for lower risk social networking, counseling of heroin injectors should aim to further reduce social links with drug-using individuals and encourage stronger links with non-IDU peers. Interventions for cocaine injectors could require alternate strategies such as individually based education and counseling sessions that nonetheless emphasize the importance of network-level risk.
This study shows that a difference in HIV and HCV infection risk exists between cocaine and heroin injectors and that this difference can be attributed to the characteristics of an IDU's social and drug-injecting networks. The identification of definable populations at risk for transmitting bloodborne infections based on drug type can be used to focus public health interventions for such high-risk individuals and their networks.
1. Single E. A Socio-Demographic Profile of Drug Users in Canada
. Ottawa, Ontario, Canada: HIV
/AIDS Prevention and Community Action Programs of Health Canada; 2000.
2. Morissette C, Roy E, Parent R, and the SurvUDI Working Group. Temporal trends in HIV
incidence and risk behaviours among IDUs in the SurvUDI study 1995-2002. Can J Infect Dis
. 2004;15(Suppl A):49A.
3. Gossop M, Marsden J, Stewart D, et al. Routes of drug administration and multiple drug misuse: regional variations among clients seeking treatment at programmes throughout England. Addiction
4. Diaz T, Chu SY, Byers RH Jr, et al. The types of drugs used by HIV
-infected injection drug users in a multistate surveillance project: implications for intervention. Am J Public Health
5. Kral AH, Bluthenthal RN, Booth RE, et al. HIV
seroprevalence among street-recruited injection drug and crack cocaine
users in 16 US municipalities. Am J Public Health
6. van Ameijden EJ, van den Hoek AR, Coutinho RA. Injecting risk behavior among drug users in Amsterdam, 1986 to 1992, and its relationship to AIDS prevention programs. Am J Public Health
7. Des Jarlais DC, Marmor M, Friedmann P, et al. HIV
incidence among injection drug users in New York City, 1992-1997: evidence for a declining epidemic. Am J Public Health
8. Chaisson RE, Bacchetti P, Osmond D, et al. Cocaine
use and HIV
infection in intravenous drug users in San Francisco. JAMA
9. Anthony JC, Vlahov D, Nelson KE, et al. New evidence on intravenous cocaine
use and the risk of infection with human immunodeficiency virus type 1. Am J Epidemiol
10. Tyndall MW, Currie S, Spittal P, et al. Intensive injection cocaine
use as the primary risk factor in the Vancouver HIV
-1 epidemic. AIDS
11. Schoenbaum EE, Hartel D, Selwyn PA, et al. Risk factors for human immunodeficiency virus infection in intravenous drug users. N Engl J Med
12. Hankins C, Alary M, Parent R, et al. SurvUDI Working Group. Continuing HIV
transmission among injection drug users in Eastern Central Canada: the SurvUDI Study, 1995 to 2000. J Acquir Immune Defic Syndr
13. Hudgins R, McCusker J, Stoddard A. Cocaine
use and risky injection and sexual behaviors. Drug Alcohol Depend
14. Bruneau J, Lamothe F, Franco E, et al. High rates of HIV
infection among injection drug users participating in needle exchange programs in Montreal: results of a cohort study. Am J Epidemiol
15. Hartel DM, Schoenbaum EE, Selwyn PA, et al. Heroin
use during methadone maintenance treatment: the importance of methadone dose and cocaine
use. Am J Public Health
16. Weiss SH. Links between cocaine
and retroviral infection. JAMA
17. Strathdee SA, Patrick DM, Currie SL, et al. Needle exchange is not enough: lessons from the Vancouver injecting drug use study. AIDS
18. Chitwood DD, McCoy CB, Comerford M. Risk behavior of intravenous cocaine
users: implications for intervention. NIDA Res Monogr
19. Donoghoe MC, Dolan KA, Stimson GV. Life-style factors and social circumstances of syringe sharing
in injecting drug users. Br J Addict
20. Compton WM, Lamb RJ, Fletcher BW. Results of the NIDA treatment demonstration grants' cocaine
workgroup: characteristics of cocaine
users and HIV
risk behaviors. Drug Alcohol Depend
21. Paone D, Des Jarlais DC, Caloir S, et al. Continued risky injection subsequent to syringe exchange use among injection drug users in New York City. AIDS Educ Prev
22. Klee H, Faugier J, Hayes C, et al. Risk reduction among injecting drug users: changes in the sharing of injecting equipment and in condom use. AIDS Care
23. Magura S, Kang SY, Nwakeze PC, et al. Temporal patterns of heroin
use among methadone patients. Subst Use Misuse
24. Bux DA, Lamb RJ, Iguchi MY. Cocaine
use and HIV
risk behavior in methadone maintenance patients. Drug Alcohol Depend
25. Joe GW, Simpson DD. HIV
risks, gender, and cocaine
use among opiate users. Drug Alcohol Depend
26. Gawin FH, Ellinwood EH Jr. Cocaine
and other stimulants. Actions, abuse, and treatment. N Engl J Med
27. Suh T, Mandell W, Latkin C, et al. Social network
characteristics and injecting HIV
-risk behaviors among street injection drug users. Drug Alcohol Depend
28. Latkin C, Mandell W, Oziemkowska M, et al. Using social network
analysis to study patterns of drug use among urban drug users at high risk for HIV
/AIDS. Drug Alcohol Depend
29. Hoffmann JP, Su SS, Pach A. Changes in network characteristics and HIV
risk behavior among injection drug users. Drug Alcohol Depend
30. Darke S, Baker A, Dixon J, et al. Drug use and HIV
risk-taking behavior among clients in methadone maintenance treatment. Drug Alcohol Depend
31. Hartel DM, Schoenbaum EE, Selwyn PA, et al. Patterns of heroin
and speedball injection among Bronx (USA) methadone maintenance patients: 1978-1988. Addict Res
32. Burt RS. Network items and the General Social Survey. Soc Networks
33. National Institute on Drug Abuse, Risk Behavior Assessment Questionnaire
. Rockville, MD: NIDA Community Research Branch; 1991.
34. Friedman SR, Chapman TF, Perlis TE, et al. Similarities and differences by race/ethnicity in changes of HIV
seroprevalence and related behaviors among drug injectors in New York City, 1991-1996. J Acquir Immune Defic Syndr
35. Brewer DD. Forgetting in the recall-based elicitation of personal and social networks. Soc Networks
36. Liang K-Y, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika
37. Maher L, Chant K, Jalaludin B, et al. Risk behaviors and antibody hepatitis B and C prevalence among injecting drug users in south-western Sydney, Australia. J Gastroenterol Hepatol
38. Rothenberg RB, Potterat JJ, Woodhouse DE, et al. Social network
dynamics and HIV
39. Neaigus A, Friedman SR, Goldstein M, et al. Using dyadic data for a network analysis of HIV
infection and risk behaviours among injecting drug users. NIDA Res Monogr
40. Curtis R, Friedman SR, Neaigus A, et al. Street-level drug markets: network structure and HIV
risk. Soc Networks
41. Valente TW, Vlahov D. Selective risk taking among needle exchange participants: implications for supplemental interventions. Am J Public Health
42. Marmor M, Des Jarlais DC, Cohen H, et al. Risk factors for infection with human immunodeficiency virus among intravenous drug abusers in New York City. AIDS
43. Neaigus A, Friedman SR, Curtis R, et al. The relevance of drug injectors' social and risk networks for understanding and preventing HIV
infection. Soc Sci Med
44. Celentano DD, Vlahov D, Cohn S, et al. Risk factors for shooting gallery use and cessation among intravenous drug users. Am J Public Health
45. Granovetter M. The strength of weak ties: a network theory revisited. In: Collins R, ed. Sociological Theory
. San Francisco, CA: Jossey-Bass; 1983:201-233.
46. Blower SM, McLean AR. Mixing ecology and epidemiology. Proc R Soc Lond B Biol Sci
47. Leri F, Stewart J, Fischer B, et al. Patterns of opioid and cocaine
co-use: a descriptive study in a Canadian sample of untreated opioid-dependent individuals. Exp Clin Psychopharmacol
48. Bedard A, Noel L, Bradet R, et al. Drug combination among untreated illicit opiate users in Quebec City site-the OPICAN cohort baseline. Can J Infect Dis
. 2004;15(Suppl A):64A.
49. Goldstein MF, Friedman SR, Neaigus A, et al. Self-reports of HIV
risk behavior by injecting drug users: are they reliable? Addiction
50. Jennings KD, Stagg V, Pallay A. Assessing support networks: stability and evidence for convergent and divergent validity. Am J Community Psychol
51. Romney AK, Weller SC. Predicting informant accuracy from patterns of recall among individuals. Soc Networks
52. Tardy CH. Social support measurement. Am J Community Psychol
53. Marsden P. The reliability of network density and composition measures. Soc Networks
54. Sudman S. Experiments in the measurement of the size of social networks. Soc Networks
55. Bernard H, Killworth P, Sailer L. Summary of research on informant accuracy in network data and the reverse of small world problem. Connections