Introduction

Among people who inject drugs (PWID), social norms are an important environmental convention which can influence individual-level HIV vulnerability and harm reduction behaviors [1,2,3,4,5,6], as posited by Bandura’s Social Cognitive Theory [7]. Alternatively, in accordance with the theory of homophily and differential association, individuals are attracted to similar others; thus, they engage in HIV risk behaviors endorsed by their social networks [8, 9]. Both theories posit that social network norms can predict future HIV risk behaviors such as sharing injection equipment (i.e., needles, cookers, or cottons; or back/front loading) [10, 11]. Peers not only establish social norms, but can also serve as a source of injection equipment, drugs, and healthy or harm reduction behaviors [10, 12]. To better understand the association between social norms and individual behavior, two types of social norms have been identified: injunctive norms, individual-level perceptions of what is acceptable by others, and descriptive norms, the actual observed behaviors within a network [13,14,15]. Much of previous research has focused on the influences of network-level injunctive norms on individual-level HIV vulnerability [1, 16, 17] and there is a dearth of research surrounding descriptive norms at the network- and individual-level, and the relationship of social norms with HIV risk behaviors among PWID [1, 2, 4, 6].

Research surrounding social network HIV risk norms among PWID has predominantly focused on PWID communities along the East Coast of the United States (U.S) [1, 2, 10, 16, 18]. However, since norms can be unique to geographic regions, information is needed outside of the context of the East Coast, in injection drug use hubs such as those along the U.S.-Mexico border [19]. Prior to the emergence of COVID-19 as a global pandemic, a large body of research established that PWID along the U.S.-Mexico border were already at elevated risk of HIV relative to other populations [20]. The San Ysidro Port of Entry separates the border between San Diego County, California, U.S. and Tijuana Municipality, Baja California, Mexico. It is the 4th busiest border crossing in the world and a major drug trafficking route [21,22,23,24]. Travel restrictions northbound at the San Ysidro Port of Entry during the COVID-19 pandemic might have exacerbated HIV vulnerability, as they prevented cross-border mobility, and disrupted established social networks [21, 22, 25, 26]. The latter may have forced PWID to join new drug use networks, potentially increasing exposure to HIV [27]. In addition, access to harm reduction services may have been affected as a result of physical distance protocols and as health system resources shifted towards the COVID-19 response [28,29,30]. Despite a political border separating these two cities, they merge to form a “melting pot” of overlapping key populations with high HIV vulnerability [31]. In 2019, HIV incidence among the general population was 13 per 100,000 in San Diego [32] and 10.3 per 100,000 in Tijuana [33]. However, HIV incidence is estimated to be much higher among PWID and preliminary research carried out after the border closure suggests that HIV incidence rate was 4-fold among a sample of PWID living in Tijuana, relative to PWID living in San Diego who engage in cross-border drug use (CBDU) [34].

CBDU and cross-border drug obtainment (CBDO) on the San Diego and Tijuana border have been increasing over the past decade [35, 36]. Drugs are generally cheaper and easier to obtain in Tijuana relative to neighboring San Diego given both drug production and supply (predominantly to the U.S.) occur in Tijuana. Differences in currency and drug policies between Mexico and the U.S., and lower exposure to police harassment in Mexico make CBDU/CBDO attractive for U.S. PWID [35, 36]. In 2009, Mexico decriminalized drug possession for personal consumption for a selection of drugs and under a specified threshold, which may also be contributing to these trends, although studies from our research group suggest limited implementation of these reforms [37]. Thus, people who engage in CBDU or CBDO are often San Diego residents who travel to Tijuana, instead of Tijuana residents who travel to San Diego, to obtain and/or use drugs. Black-tar heroin and methamphetamine use, often in combination, have been consistently reported in both Tijuana and San Diego over the past decade [38,39,40,41]. However, the picture has been changing recently, as cartels have shifted from black tar to fentanyl production and supply as this is much more profitable than heroin [42]. California experienced one of the highest increases (> 45%) in overdose deaths across the U.S. between 2020 and 2021 [43, 44]. Over this same period, there was a 61% increase in overdose deaths in San Diego County (871 overdose versus 528), and ≥ 60% were fentanyl-related [45]. In 2022, 16% of PWID in San Diego had an overdose in the prior 6 months, which has doubled since 2014 [46, 47]. Between 2011 and 2013 in Tijuana, 17.4% of PWID had an overdose within the last 6 months [48]. Overdose deaths are poorly monitored in Tijuana but increases in the use of powder heroin and fentanyl presence in drug samples along the Mexican side of the border have been confirmed [49]. In addition to increasing fatal overdose risk, fentanyl use is associated with greater risk of HIV infection as, unlike black-tar heroin use, its use does not require heating or rinsing of syringes in order to prevent clogging, a harm reduction strategy to inactivate HIV [50,51,52]. Additionally, fentanyl’s short half-life has also been shown to lead to more frequent injecting and by extension, to more frequent syringe sharing [50, 53].

A study of PWID living in Mexico found that nearly 20% of PWID in Tijuana recently injected with U.S. PWID who engaged in CBDU, suggesting that CBDU can enable cross-border HIV transmission [54]. While historically, cross-border HIV transmission primarily occurred from the U.S. to Mexico, for the first time, cross-border transmission is now occurring from Mexico to the U.S. [31]. Consequently, CBDU and CBDO as a phenomenon warrants further investigation. Harm reduction strategies to reduce HIV vulnerability from injection drug use have been implemented differentially across the two countries [55, 56]. The uptake of harm reduction services is not only impacted by service availability, but also may also be influenced by social network norms. Access restrictions to harm reduction strategies during the COVID-19 pandemic [26] and previous needle sharing norms might have exacerbated HIV risk among PWID who live along the U.S.-Mexico border. It is therefore important to characterize social network norms relevant to specific communities highly vulnerable to HIV, as norms are diverse across different environments, continuously evolve, and rapidly disseminate throughout networks [4].

The aim of the present study was to examine the influence of PWID’s descriptive network norms from a person-centered perspective on HIV-related risk and harm reduction behaviors during the COVID-19 pandemic (post-implementation of border crossing restrictions) [26, 57]. We used Latent Profile Analysis (LPA) to categorize PWID into person-centered descriptive network risk norms profiles. LPA was used due to its ability to identify unique subgroups based on social network norms that variable-centered analyses alone cannot elucidate. We examined the relationship between descriptive network risk-norms profiles and individual-level HIV risk and harm reduction behaviors during the previous 6 months.

Methods

Setting and sample description- The parent study, La Frontera, is a longitudinal study among PWID in which researchers aim to characterize trends in incidence of HIV, Viral Hepatitis C and drug overdose associated with binational drug markets and CBDU between San Diego, US, and Tijuana, Mexico. Data for the present study were baseline and supplemental visit data of La Frontera. Additional study information can be found elsewhere [58]. Participants were PWID aged ≥ 18 from 3 groups: (1) PWID who live in San Diego but engaged in CBDU in Tijuana in the past 2 years, (2) PWID who live in San Diego and had not been to Mexico in at least 2 years or (3) PWID who live in Tijuana and had not been to the U.S. in at least 2 years. We operationalized PWID who did not engage in CBDU as PWID who have not used illicit drugs across the border from where they reside in the past 2 years. Participants were recruited through street outreach between October 2020 and October 2021. Trained interviewers obtained written informed consent then administered a computer-assisted sociodemographic and behavioral questionnaire. Within two weeks of the initial visit, interviewers administered a computer-assisted supplemental questionnaire which assessed participants’ social network and HIV-related risk factors.

Sociodemographic characteristics- We collected data on participants’ sociodemographic characteristics such as age, ethnic identity (Non-Hispanic or Hispanic), race (Black, White, Mixed, Asian, Native American, or other), sex assigned at birth (male or female), and country of birth (U.S., Mexico, or other).

HIV status and serology- Samples underwent HIV serology at baseline. Rapid HIV tests were conducted using the Miriad® HIV Antibody InTec Rapid Anti-HCV Test (Avantor, Radnor, PA) [59]. Reactive and indeterminate tests underwent a second rapid test with Oraquick® HIV (Orasure, Bethlehem, PA) [60].

Network risk variables- Social network information was collected by asking participants to name up to 20 people they had seen in the past 30 days and who they talked to about things that are important to them (alters). Network risk information was collected for the first five alters that participants named. Only alters that used drugs were included for analyses. Based on previously established HIV risk factors, participants indicated whether each alter (1) used drugs by injection or (2) by non-injection (i.e., “Does [ALTER] use drugs by injection, by non-injection, or sometimes one and sometimes the other?”), (3) lived in Mexico at the time of the study (i.e., “Where does [ALTER] live?”), (4) ever crossed the border to buy or obtain drugs (i.e., “In the past 6 months, had [ALTER] traveled to Mexico/U.S. to buy or obtain drugs?” i.e., CBDO), (5) shared injection equipment with the participant (i.e., “Have you ever used a needle, water, cooker or cotton that had already been used by [ALTER]?”), (6) offered to share drugs with the participant (i.e., “Has [ALTER] ever offered to share or encouraged you to use drugs?”), and (7) either used more than or double their usual dose of drugs (i.e., “Consider [ALTER] usual daily drug use, does [ALTER] ever use double the amount or mix with other drugs?”). Then, for each network risk variable we calculated the proportion of participants’ drug use networks that engaged in the network risk variable.

HIV behavioral risk and harm reduction variables- Participants’ HIV risk was assessed through self-reports of 4 risk behaviors and 2 harm reduction behaviors. Risk behaviors in the past 6 months included consistency of (1) giving, renting, or lending a syringe the participant already used to someone else; (2) using a syringe that participant knew or suspected had been used before by someone else; (3) dividing up drugs with somebody else by using a syringe (i.e., back loading, piggybacking or splitting drugs wet); and/or (4) using a cooker, cotton, or water with someone or after someone else. Responses were collected in Likert scale form (1 = never; 2 = less than half the time; 3 = about half of the time; 4 = more than half the time; and 5 = always). We created an index to measure HIV risk, which was calculated by finding the mean score of participants’ responses to the aforementioned items. The index had excellent internal consistency (Cronbach’s α = 0.908).

HIV harm reduction behaviors included (1) consistency of injecting with a new, sterile syringe in the past 6 months and (2) having ever been tested for HIV. Injecting with a new, sterile syringe was assessed as a Likert scale for consistency in the past 6 months (1 = never; 2 = less than half the time; 3 = about half of the time; 4 = more than half the time; and 5 = always). Having ever been tested for HIV was measured dichotomously (1 = no; 2 = yes).

Statistical analysis- LPA was used to categorize PWID into empirically-based network risk-norms profiles based on the proportion of their network which engaged in specific risk behaviors, using a person-centered approach [57]. The indicators for LPA included information about alters who used injection and non-injection drugs, engaged in CBDO, lived in Mexico, shared a needle with the participant, offered the participant drugs, and either doubled their daily use or mixed drugs, as mentioned above. The network risk-norms latent variables were itemized and constructed as continuous measures based on the proportion of their network who engaged in each behavior, with higher values indicating higher network norm of that variable.

The outcomes of interest were the HIV risk index, and the two harm reduction variables. To assess significant differences between network risk-norms profiles by sociodemographic and HIV risk and harm reduction behaviors, we used chi-square goodness-of-fit tests and Analysis of Variance (ANOVA). Lastly, we identified associations profile membership using post-hoc linear and binomial logistic regression to identify associations with HIV behavioral risk and harm reduction outcomes. We tested five assumptions of linear regression: linearity, homoscedasticity, multicollinearity, independence, and normality [61]. Basic assumptions that were met for logistic regression include independence of errors, linearity in the logit for continuous variables, absence of multicollinearity, and lack of strongly influential outliers [62]. We observed the distribution and outliers of our data through visualizations (i.e., histogram) and by measuring the skewness and kurtosis of each numerical variable. Additionally, we looked at the correlation of variables and the VIF of models. All assumptions of linear and logistic regressions were met. The R environment was used to conduct the LPA (tidyLPA) and the regressions (lm for linear regression and glm for logistic regression).

The University of California San Diego ethics committee provided ethical approval for this study.

Results

Of 612 participants who were recruited and enrolled in the parent longitudinal study, 399 PWID (n = 150 San Diego residents who engage in CBDU, n = 90 San Diego residents who did not engage in CBDU, n = 159 Tijuana residents who did not engage in CBDU; 65% of entire sample) provided additional social network data (n = 1,226 alters) and only those network members who participants indicated drug use for in the past year (n = 924 alters who used drugs within the past year) were included in the LPA. Participants’ mean age was 44 years and 26% were assigned female sex at birth. All participants identified as cis gender. Additional information about participants can be found in Table 1, by latent profile.

Additional information about CBDU group

Among the group of participants who engaged in CBDU (n = 150), participants were either interviewed in San Diego or Tijuana. Participants who were interviewed in San Diego reported having gone to Mexico a mean of 2.6 months ago. Participants interviewed in Tijuana reported that the previous time that they travelled to Mexico before the day of the interview was a mean of 1.1 months ago. Participants reported going to Mexico a mean of 30.7 times in the last 6 months (min = 1; max = 180 times). The mean length of time that participants reported staying in Tijuana was 16 days (min = 0.05, max = 279 days). Of participants who engaged in CBDU, 90% reported using or obtaining drugs or drug paraphernalia, 49% reported visiting friends, and 37% reported visiting family. Of participants who reported CBDU, the most commonly used drugs in Mexico included heroin (91%; of which 69% reported using primarily black tar heroin), methamphetamine or crystal meth (71%), marijuana (56%), and the combined use of heroin and methamphetamine or crystal meth (43%). Additional information about participants who engaged in CBDU can be found in Appendix 1.

Table 1 Associations between network HIV risk norms profile and background characteristics
Full size table

Selection of latent profile model of most parsimonious fit

To assess which model of the five profiles best fit the data, we considered several fit criteria as described in in Table 2 [63]. We first assessed Akaike’s information criterion (AIC) and Bayes information criterion (BIC) for lower values as these indicate a more parsimonious fit of the data [64,65,66]. We then considered the models with an Entropy value greater than 0.8 as this indicates the model’s ability to discriminate between profiles [67]. In considering AIC, BIC, and Entropy, the four- and five-profile models fit the data best. We then considered the Lo-Mendell-Rubin likelihood ratio test (LMR-LRT) and Bootstrap likelihood ratio test (BSLRT) as this indicates if a given model fits the data better than the model with one profile less [68, 69]. The BSLRT suggested that only the six-profile model did not fit the data better than the model with one profile less (BSLRT value = 316.92; p = 0.99). Lastly, we considered the smallest sample size of each profile, as models which include profiles that are less than or equal to 25 members could indicate a spurious profile [70]. The Prob. Min (minimum of the diagonal of the average latent profile probabilities for most likely profile membership, by assigned profile) and Prob. Max (maximum of the diagonal of the average latent profile probabilities for most likely profile membership, by assigned profile) signify greater classification as values increase, and should be as high as possible [71]. For this, we chose the four-profile model, which had the most parsimonious fit to the data. Figure 1 visualizes the mean probability of being included in social norm risk profiles for each network risk variable.

Table 2 Fit indices for network norm HIV risk latent profile analysis
Full size table

Network norm latent profile profile description

As the fit indices suggested a 4-latent profile solution, network norm profiles were classified as (1) lower risk network norm profile (n = 178), (2) moderate risk with CBDO network norm profile (n = 34), moderate risk without CBDO network norm profile (n = 94), and 4) higher risk network norm profile (n = 93), as observed in Fig. 1. Figure 1 demonstrates that higher risk network norm profiles have networks in which the majority of alters participated in high HIV risk behaviors. The moderate risk network norm profiles were composed of network with alters who had similar HIV risk behaviors, with two exceptions: (1) proportion of alters who engaged in CBDO and (2) proportion of alters who lived in Mexico. The lower risk network norm risk profile was composed of networks in which the majority of alters did not practice high HIV vulnerability behaviors. Table 1 describes sociodemographic characteristics of participants, stratified by network norm HIV risk profile membership, with significant relations examined by bivariate associations. There were differences in profile membership based on Hispanic ethnicity, race and ethnicity, and birth country.

Fig. 1
figure 1

Four profile LPA plot of network HIV risk norms within the social networks of PWID

Full size image

Bivariate associations between network norm profile and HIV risk and harm reduction behaviors

There were statistically significant differences between network norm profiles and study groups, our HIV risk index, and the harm reduction strategies examined. There were significant differences in distribution of profile membership by network norm profiles (χ2 = 63.36, p < 0.01). The lower risk network norm profile consisted of an equal distribution of study groups (~ 33%), the moderate risk with CBDO network norm profile consisted of a majority of participants who engaged in CBDU (81%), the moderate risk without CBDO network profile consisted of approximately equal distributions of participants who engaged in CBDU (44%) and Tijuana group (40%), and the higher risk network norm profile consisted of a majority of participants from the Tijuana group (71%). PWID in the higher network-risk norms profile reported more HIV vulnerability behaviors relative to the lower network-risk norms profile (2.59 vs. 1.79; F-value: 12.8, p < 0.001) and PWID in the two moderate network risk norms profile reported a similar number of behaviors (2.01 vs. 1.97). Interestingly, a higher proportion of PWID in both the higher (74.2%) and the lower (76.8%) network risk norms profiles had ever tested for HIV compared to the two moderate risk profiles (with CBDO = 47.1%; without CBDO = 61.7%; χ2 = 16.35, p < 0.001). Mean Likert scale value of consistency of injecting with a new, sterile syringe in the past 6 months varied by network profile, with the lower network risk norm profile reporting the highest consistency (mean = 4) and higher network risk norm reporting the lowest consistency (mean = 3.51; F-value = 3.97, p < 0.001).

Adjusted multivariate linear regression: HIV risk outcome

Relative to participants in the lower risk network norm profile, the HIV risk index score increased for participants in the moderate risk (with CBDO) network norm profile (β = 0.4; 95% CI: 0.02–0.78; p = 0.04) and higher risk network norm profile (β = 0.65; 95% CI: 0.38–0.91; p < 0.001). In addition, the HIV risk index score increased for those in the Tijuana study group who did not engage in CBDO (β = 0.64; 95% CI: 0.40–0.89; p < 0.001) relative to the CBDU study group. Full model specifications can be found in Table 3: Model A.

Adjusted multivariate linear regression: injected with a new syringe harm reduction outcome

Participants in the moderate risk (with CBDO) network norm profile (β=-0.47; 95% CI: -0.93 – -0.02; p = 0.041), moderate risk (without CBDO) network norm profile (β=-0.36; 95% CI: -0.66 – -0.06; p = 0.019), and higher risk network norm profile (β=-0.48; 95% CI: -0.79 – -0.17; p = 0.003), relative to participants in the lower risk network norm profile, had significantly lower consistency of injecting with a new, sterile syringe. Participants who were in the Tijuana group (relative to CBDU; β=-0.41; 95% CI: -0.70 – -0.12; p = 0.006) also experienced decreased consistency in having used sterile syringes. However, participants who identified as Hispanic (relative to not Hispanic; β = 0.40; 95% CI: 0.08–0.72; p = 0.015) were more likely to inject with a new, sterile syringe in the past 6 months. Full model specifications are included in Table 3: Model B.

Results of adjusted multivariate logistic regression: ever tested for HIV harm reduction outcome

Participants who were in the moderate risk with CBDO network norm profile (OR = 0.36; 95% CI: 0.16–0.80; p = 0.013) and in the moderate risk without CBDO network norm profile (OR = 0.56; 95% CI: 0.32–1.00; p = 0.048) were less likely than participants in the lower risk network norm profile to have ever been tested for HIV. Relative to participants who engaged in CBDU, participants who were in the San Diego study group (OR = 3.17; 95% CI: 1.59–6.61; p = 0.001) were more likely to have had an HIV test. Those who reported more alters (OR = 1.21; 95% CI: 1.01–1.45; p = 0.036) were also more likely to have had an HIV test. Additional model specifications are included in Table 3: Model C.

Table 3 Multivariate adjusted regression models examining associations between network norm profiles and HIV risk and harm reduction behaviors
Full size table

Discussion

This study aimed to identify descriptive network-level risk norm profiles among three PWID groups who live along the US-Mexico border, and then identify associations between these profiles and individual-level HIV risk and harm reduction behaviors. We found that network descriptive norms could be categorized into distinct profiles and these profiles were associated with HIV risk and harm reduction behaviors. We identified four distinct profiles which described network norms: a lower risk network norm profile, a moderate risk network norm profile composed of alters who engaged in CBDO, a moderate risk network norm profile composed of alters who did not engage in CBDO, and a higher risk network norm profile. We found that PWID who were classified into moderate and higher risk network norm profiles (relative to the lower risk network norm profile) were significantly more likely to engage in individual-level HIV risk behaviors and significantly less likely to engage in HIV harm reduction behaviors.

Our findings suggest that CBDU, CBDO, and place of residency play a role in access to harm reduction strategies. For example, Tijuana-based PWID may have reduced access to sterile syringes as they are less likely to inject with a clean syringe relative to San Diego-based PWID who engage in CBDU. In addition, PWID who live in San Diego and do not engage in CBDU were more likely to have received an HIV test relative to PWID who live in San Diego and engaged in CBDU. These service gaps could have been compounded by the COVID-19 pandemic, which resulted in further cuts to the already meagre and irregular harm reduction budget in Tijuana and the disruption of health and harm reduction services in San Diego [29, 30]. The moderate risk profile with CBDO engaged in higher risk behaviors that the other moderate risk profile without CBDO and the lower risk profile, indicating that socializing with people who engage in CBDU is associated with higher individual-level HIV risk behaviors. Participants in the CBDU study group were defined as those who reported being a resident of San Diego who crossed the border to inject drugs in Tijuana within the 2 years prior to baseline. Thus, participants were different from their alters in the sense that participants did not specifically travel across the border to use, buy, or obtain drugs- they may have crossed the border for different reasons such as visiting family and then engaged in CBDU because they had already crossed the border into Mexico. If PWID who do not engage in CBDU have exposure to CBDU and CBDO norms and behaviors, they may have increased exposure to HIV within these drug use networks. PWID who engage in CBDU and CBDO may serve as a bridge between injection networks. Additionally access restrictions to harm reduction strategies and previous syringe sharing norms likely exacerbated HIV risk among PWID in Tijuana and our analysis indicates that engaging in CBDU and having people who engage in CBDU or CBDO in one’s social network was associated with higher risk behaviors, and potentially amplified HIV vulnerability. Thus, interventions should take a social network approach and strive to change the network’s norms of injection drug use.

Social network interventions provide an opportunity to decrease HIV risk network norms, decrease individual-level risk behaviors, and increase individual-level harm reduction behaviors [43]. A systematic review published in 2017 identified 58 studies which tested various social network strategies to increase HIV prevention among people who use substances [72]. Relevant to our priority population of PWID, social network interventions can leverage social diffusion interventions and peer change agents; a peer change agent intervention could include PWID peers referring PWID to harm reduction services [18]. Another opportunity to intervene could be to distribute and facilitate access to syringes in settings such as Tijuana where harm reduction services are insufficient, but syringes are sold at pharmacies. Peer educators may also influence social norms by engaging in HIV harm reduction behaviors which can potentially change the descriptive norms or communication norms of a social network [18]. Bouchard et al. (2018) found that although a network may be saturated with people who practice some harm reduction behaviors, only a minority of networks consisted of PWID who were champions of harm reduction strategies [12]. We also found that more expansive social networks (i.e., higher number of alters named) was associated with ever testing for HIV, a harm reduction behavior. This suggests that increasing PWID’s social network characteristics, such as network size could increase exposure to harm reduction services within networks and serve as an intervention component.

For social network interventions to be effective in eliciting the desired behavioral change (i.e., reducing HIV risk behaviors and increasing harm reduction behaviors), conversations must highlight network-level health promotion social norms so individuals are aware of their social network member’s health behaviors [10]. Future interventions can consider including CBDU or CBDO reduction components; as networks with people who engage in CBDO may increase individual-level HIV vulnerability, or focusing on reducing risk behaviors among people who engage in CBDU, as these may be influential members in networks on both sides of the border. Previous research suggests that reducing the number of people within a network who inject could also decrease drug use by injection [73]. Another social network intervention could be to train PWID with large networks in buying syringes for themselves and others within their networks. Future studies and intervention research should also consider the multilevel and larger social contexts that PWID live in: violence, fear of withdrawal, and fear of police, among other issues, could serve as barriers to the harm reduction strategy of needle exchange.

Our study is unique because previous studies focus on injunctive norms instead of descriptive norms [1, 2, 11, 74, 75]. Our operationalization of social norms focuses on descriptive norms, which are observed, instead of injunctive norms, which are perceived. In addition, those studies that did include measures of descriptive norms did not use LPA, which can be powerful in identifying person-centered patterns of risk [1, 17]. However, our study is not without limitations. Our first limitation is the lack of social network data collected. Of a total of 612 participants, only 399 were included in our secondary analysis due to a lack of participant familiarity with network data collection, participants’ loyalty and protection of their social networks, and potential mistrust with researchers. Some participants clearly stated during assessments that they would not share their network information with our study team. This may be due to police falsely impersonating researchers to gain access to drug use related information: in our San Diego site, study participants relayed stories of undercover police presence on the streets and infiltrating their networks. Significant police presence was observed in the surroundings of some of the recruitment spots in the community. However, addressing fear of police could be beneficial for harm reduction interventions. Previous interventions found that law enforcement officers could provide syringe exchange site referrals to PWID [76]; thus, improving the relationships between law enforcements officers and PWID could facilitate better data for future studies. Second, due to the retrospective self-report items in our questionnaire, we may have introduced recall bias. In addition, despite our data being longitudinal, we opted to use cross-sectional data for the present analyses and we did not examine change in risk profile or behaviors relative to prior to the COVID-19 related border closure. A future study will incorporate a longitudinal approach. Finally, we used an egocentric approach. A sociocentric approach, which includes the recruitment of an entire network, could have made our approach stronger. Examples of sociometric networks would include groups of people who inject together or who may engage in CBDU or CBDO together.

Conclusions

PWID had a gradient of HIV risk within their networks, based on network norms. PWID in social networks in which descriptive norms included higher risk HIV behaviors, such as CBDU/CBDO social norms, were more likely to engage in behaviors which increased behavioral vulnerability to HIV. Longitudinal research is needed to understand long term effects of border closure on network risk norms and HIV risk outcomes. Interventions should focus on reducing HIV risk among PWID with higher risk networks, particularly when services are limited and networks are PWID’s main source of influence.