This study was a cross-sectional secondary analysis of prospective data from a continuously enrolled population of patients presenting with injuries to the Kenyatta National Hospital (KNH) ED in Nairobi, Kenya. Data was collected from March to August 2023 and derived from a study assessing HIV care delivery in this setting.48 KNH is the largest hospital in Kenya, and serves as the primary injury center for Nairobi and a national referral hub. The KNH ED provides 24-hour emergency services and has embedded HIV Testing Services freely available day and night.495051

Patients were recruited from all individuals seeking ED care; eligible patients were adults aged 18 years or older, obtaining injury care at the KNH ED and able to provide informed consent. Exclusions included non-injury cases, incarcerated persons, pregnant individuals, and critically ill patients or those unable to consent. All enrolled participants provided informed consent, with study information and consents available in Kiswahili and English, and research staff were fluent in both languages. The study was approved by both the KNH Ethics and Research Committee and the Institutional Review Board of Rhode Island Hospital.

Participants received standard ED care for injuries from non-study providers. Screening and consent were conducted in a private area to ensure confidentiality, with enrollment occurring as soon after triage as able. Data collected included demographics, medical history, social history injury details, interpersonal violence history, and ED care outcomes.49 All data were securely stored in electronic databases using password-protected devices accessible only to study staff.5253

Empirical approaches to the SAVA theory have relied on steps that require simplification of complex interactions and can fail to capture the nuanced interactions between substance abuse, violence, and HIV/AIDS.54 A prior scoping review of SAVA research revealed that the majority of studies relied on regression analyses rather than employing other methods, that could be superior in evaluating the complex inter-relationships.17

Demographic, HIV risk, substance use, and violence risk factors were utilized for modeling and analysis. Data for all variables were based on participant self report. Demographic variables included age at time of ED visit and biological sex. HIV risk factors included sexual identity (gay and/or MSM), sex partners HIV status, condom use practices during the last six months, history of injecting drug use, transactional sex (money or other goods exchanged to give or receive sex), and following the Kenya national guidelines on recommend HIV testing frequency (if age 18-24 years testing every 6 months, if age ≥25 years and no other risk factors testing every 2 years, if high risk ≥18 years testing every 3 months).55 Substance use risk factors included: illicit drug use during the last 6 months, drug use during the time of injury, history of polysubstance use, and being at risk for hazardous alcohol use based on the validated Alcohol Use Disorders Identification Test Consumption (AUDIT-C) questions (females scoring ≥3 and males ≥4).565758 Violence exposure risk factors included report of either emotional, physical or sexual violence in the past 6 months.4759

Data was imported into SAS (Version 9.4, SAS Institute Inc, Cary, N.C.) for analysis. Descriptive statistics are reported as means with standard deviations (SD), medians with interquartile range (IQR), or percentages where appropriate. Prior to modeling, data used in the analysis were transformed into binary values (age 18-24 versus older, yes or no to substance use variables, positive or negative for AUDIT-C, HIV risk present or not, etc. Due to low frequencies for variables defining key populations including ever injecting drugs (n = 4), MSM (n = 5), gay (n = 4), and commercial sex work (n = 9) a summary composite ordinal variable was derived and used in modeling analyses for HIV risk.

Structural equation modeling (SEM) was used to identify measured indicator variables associated with three latent factors; HIV risk, violence exposure, and substance use risk. SEM is a statistical approach used to evaluate relationships between observed and latent (unobserved) variables to better understand direct and indirect effects within a system of relationships.6061 The SEM method has been used in a variety of other settings62636465 , however, this methodology has not previously been utilized to evaluate the SAVA syndemic in emergency care settings. SEM combines factor analysis (to define the underlying latent factors from their measured manifest variables) and predictive models (assessing the covariance between the latent constructs). Model fitness was evaluated by assessing the Bentler-Bonett Normalized (BB NFI) and Goodness of Fit Indices (GFI, acceptable values >0.90) and the standardized root mean square residual (SRMSR, acceptable values <0.05). The covariance between the proposed three latent syndemic factors was also estimated.

An exploratory and confirmatory factor analysis was conducted, a random sample of 400 participant data was selected for the exploratory model, and the full sample used in the confirmatory model. As binary variables were being used in these analyses, the diagonally weighted least squares (DWLS) estimation method was used. Following the SEM analysis a mediation model was added to assess the potential mediating effect of key populations for HIV on the relationship between the identified latent risk factors.66

During the study period 2368 patients presenting to the KNH ED were screened for participation. Among those, 1078 (45.5%) did not meet inclusion criteria and 625 (26.4%) declined participation. A total of 665 patients were consented and enrolled (Figure 1). Males were 83% of the sample and the mean age was 32 years (standard deviation of 10.5 years). Patient characteristics regarding SAVA syndemic factors are summarized in Table 1. A number of findings suggest increased risks for syndemic SAVA conditions among enrollees, including: not knowing sexual partner(s) HIV status (89.6%), high levels of illicit drug use during the injury event that led to the ED encounter (18.5%), and high rates of prior 6-month experiences of physical violence (18.9%).

The exploratory and confirmatory SEM yielded two identified latent constructs that were supported by the overall fit indices and beta coefficients of the individual variables on the factor (Figure 2). The overall fit indices indicated a close fitting two-factor model (GFI=0.98, BB NFI=0.95, SRMR=0.04). The indicator variables had significant loadings on two of the three syndemic variables, substance use risk and violence exposure (ranging 0.25 to 0.95), but only MSM had a significant factor loading on the putative HIV risk factor, insufficient to determine that the latent factor had been identified. Additionally, violence exposure and substance use risk had significant and positive covariance (β=0.30, p<0.001).

Following the SEM, a mediation path analysis was conducted with the direct path between substance use risk and violence risk compared to a mediated pathway between substance use risk and violence risk when key populations the mediating variable (Figure 3). Fit indices for this were within acceptable ranges (GFI=0.92, BB NFI=0.94, SRMR=0.05). Illustrated in Figure 3 the direct path between substance use risk is significant and positive in the direct path (c=0.24, p<0.001). The inclusion of the mediation variable, the key population composite HIV risk, partially mediated the relationship between substance use and violence. The direct path remained significant (c’=0.22, p < 0.001) and the indirect paths between substance use and key populations (a=0.15, p < 0.001) and key populations and violence risk (b = 0.08, p < 0.05) were also significant (Figure 3).

ARA and the overall work was supported by the National Institute of Allergy and Infectious Diseases (grant number: K23AI145411). Disclaimer: The contents of this manuscript are solely the responsibility of the authors and do not necessarily represent the official views of affiliated institutions or funding bodies.

The study was approved by both the Kenyatta National Hospital (Nairobi, Kenya) Ethics and Research Committee and the Institutional Review Board of Rhode Island Hospital (Providence, RI, USA).

All participants provided informed consent, with study information and consents available in Kiswahili and English

The data that support the findings of this study are available from the corresponding author, JAL, upon reasonable request.