Risk factors for infective endocarditis and serious injection related infections among people who inject drugs in Los Angeles, CA and Denver, CO

Abstract

Background:

Injection drug use-related infective endocarditis (IDU-IE) and bacterial infections have grown in the United States, but little is known about risk factors for these infections in community samples of people who inject drugs (PWID).

Methods:

During 2021–22, PWID were recruited from community settings and surveyed for history of IDU-IE, serious injection related symptoms (SIRI) and untreated infection symptoms in the last 3 months. We used bivariate analysis and multiple logistic regression to examine factors associated with these outcomes.

Results:

Among participants (n=472), 7% reported ever having IDU-IE, 14% reported having SIRI symptoms and 20% reported untreated infection symptoms in the last 3 months. Ever having IDU-IE was associated with HCV (adjusted odds ratio [AOR]=8.37; 95% confidence interval [CI]=2.46, 28.49), prior MRSA infection (AOR=5.37; 95% CI=2.44, 11.80), identifying as female and/or gender minority person (AOR=3.14; 95% CI=1.42, 6.95). SIRI symptoms were associated with greater material hardship (compared to low; AOR=2.47; 95% CI=1.17, 5.22), fentanyl use (AOR=2.15; 95% CI=1.01, 4.61), sharing filter/cotton (AOR=1.93; 95% CI=1.10, 3.39), and licking needle prior to injection (AOR=1.85; 95% CI=1.02, 3.36). Untreated infection symptoms were associated with poor quality sleep (AOR=2.04; 95% CI=1.21, 3.43), any mental health diagnoses (AOR=2.01; 95% CI=3.56), any chronic pain (AOR=1.89; 95% CI=1.14, 3.11), sharing filters (AOR=1.81; 95% CI=1.10, 2.98), and prior MRSA infection (AOR=1.75; 95% CI=1.04, 2.97).

Conclusion:

Risk factors identified include treatable co-morbidities (i.e., HCV & MRSA history, mental health, pain, opioid use), modifiable health behaviors (i.e., equipment sharing, needle-licking), and addressable structural conditions (material hardship, housing).

1. Introduction:

Overdose death rates have risen dramatically in the last decade in the United States (Centers for Disease Control and Prevention, 2023). More than 106,000 persons in the U.S. died from drug-involved overdose in 2021, including illicit drugs and prescription opioids (Spencer et al., 2022). The United States has also seen a rise in bacterial infections and injection related infectious diseases. (Bates et al., 2019; Chandrasekar & Narula, 1986; Collier et al., 2018; Fleischauer et al., 2017; Kadri et al., 2019; McCarthy et al., 2020; Phillips & Stein, 2010; Ronan & Herzig, 2016)s. Potential drivers include co-occurring nationwide increases in rates of injection drug use (Jones et al., 2017; Bradley et al., 2020) and unsheltered houselessness (Henry et al., 2021). Community acquired infections, water, hygiene, and sanitation (WaSH) insecurity, and material hardship, are disproportionately higher among unhoused people who use drugs, and can contribute to negative health outcomes, including serious injection related infections (SIRI) (Avelar Portillo et al., 2024; Avelar Portillo et al., 2023; Goldshear et al., 2022; Packer et al., 2019). Additionally, lack of access to harm reduction services such as sterile syringes and injection equipment, fear of criminalization, and widespread stigmatization of PWID contribute to unsafe injection and associated morbidity and mortality (Barocas et al., 2021a; Bluthenthal et al., 2024; Flath et al., 2017; Ganesh et al., 2024). Injection drug use related infections range from methicillin-resistant staphylococcus aureus (MRSA) to SIRIs such as abscesses, systemic infections such as injection drug use related infective endocarditis (IDU-IE), AA amyloidosis, osteomyelitis, and sepsis (M. Harris et al., 2018; Hrycko et al., 2022).

Population based data have shown an increase in IDU-related infections. Endocarditis, osteomyelitis, sepsis, and SSTI hospitalizations increased by 33%, 35%, 24%, and 12%, respectively between 2016 and 2018 (Coyle et al., 2020). In US based hospitals, patients with Substance Use Disorder (SUD) showed increased hospitalizations for IE (1.1 to 2.1 per 100 000 persons), osteomyelitis (1.4 to 2.4 per 100 000 persons), central nervous system abscesses (0.5 to 0.9 per 100 000), and SSTIs (4.4 to 32.9 per 100 000 persons) (McCarthy et al., 2020). Between 2012 to 2017, for patients between 18–44 years, substance use related IE hospitalizations more than doubled and 50.3% of patients with substance use related IE hospitalizations had a Staphylococcus aureus infection (McCarthy et al., 2020).

While dramatic increases in IDU-IE and other SIRI have been noted in the U.S. (Wurcel et al., 2016) and are likely to continue to increase without adequate access to care and harm reduction services (Barocas et al., 2021a), little is known about sociodemographic and drug-related risk factors for these infections in community samples of people who use drugs (Kadri et al., 2019; McCarthy et al., 2020). We sought to examine whether demographic, drug use, measures of subsistence (Gelberg et al., 1997), and injection-related risk behaviors were associated with a history of IDU-IE, untreated infection symptoms, and SIRI symptoms among community-recruited people who inject drugs (PWID).

2. Methods:

2.1. Study sample:

Data for this analysis come from the study “Cannabis use and health outcomes among opioid-using people who inject drugs in the context of cannabis legalization” (NIDA R01DA046049). In this study, PWID were recruited from three community sites in Los Angeles, CA and two sites in Denver, CO (include syringe service programs, homeless service agencies, and community settings frequented by PWID) between April 2021 and November 2022. Here, we analyzed data from the baseline interviews of the study (n=472).

Study inclusion criteria were self-reported injection drug use (confirmed by visual inspection of injection stigmata or “track” marks) (Cagle et al., 2002), opioid use in the last 30 days, being 18 years of age or older, and having the capacity to complete informed consent procedures. Physical examinations were not a part of study procedures. It is important to note that it was not necessary for participants to be injecting opioids. For example, a participant who was smoking fentanyl but injecting another substance would qualify for the study.

Participants completed a 45-minute computer-assisted personal interview using the Questionnaire Design System (NOVA Research, Bethesda, MD). The survey included a total of 351 items. Sections of the questionnaires are as follows: A) Background, B) Substance use, C) Syringe access and injection practices, D) Overdose, E) Withdrawal, F) Sexual behavior, G) Health, H) Violence, I) Law enforcement contact, J) Drug treatment, and K) Sleep and subsistence. They received $20 for completing the baseline interview. All study procedures were reviewed and approved by the institutional review board at the University of Southern California (approval number HS-18–00624).

2.2. Key Variables:

We examined three outcome variables, ever having IDU-IE, symptoms related to serious injection related infection (SIRI) and symptoms of untreated, non-specific infections. First, we asked participants: “Have you ever been diagnosed with infective endocarditis (an infection to your heart valve)?” We categorized those responding, ‘yes’ as ‘ever having IDU-IE’. We categorized all others as not ever having IDU-IE. For symptom assessments, we asked participants: “In the past 3 months, have you experienced high fevers, shaking chills, night sweats or other symptoms due to a potentially untreated infection?” We categorized ‘yes’ responses as having an untreated, non-specific infection, recoded all other responses to ‘no.’ We asked affirmative responders, “Were those symptoms accompanied by increased shortness of breath, weight loss, back pain, chest pain, bloody urine, painful spots on your fingers or toes, swelling in your legs, loss of or changes in vision, or new headaches?” ‘Yes’ responses were categorized as having ‘serious injection related infection symptoms’, and we recoded all other responses to ‘no.’ Despite asking about Janeway lesions or Osler nodes (described in the survey item as painful spots on your fingers or toes) which are typically associated with infective endocarditis we did not classify these signs as IDU-IE specific. This is because participants may be responding to some or all symptoms and there may be other invasive infections including but not limited to osteomyelitis and epidural abscesses, that may cause PWID to report symptoms such as increased shortness of breath, weight loss, back pain, and chest pain. Additionally, while there are clinical gold standards for detecting specific infections such as an echocardiogram for infective endocarditis, this study used only self-reported measures as we recruited participants within community settings. Hence, we are unable to distinguish these symptoms from an untreated bacteremia in this population. In this way, ‘yes’ responses to this item most likely indicate that PWID have experienced symptoms consistent with potentially untreated invasive bacterial and fungal infections related to injection drug use in the last 3 months. We classified them as “SIRI symptoms” and hereafter, refer to them accordingly. Finally, ‘untreated, non-specific infection symptoms’ were those that could present for myriads of diseases including but not limited to invasive infections and/or respiratory conditions common among similar samples of PWID.

We selected potential predictor variables based on prior studies and biological plausibility, the specific variables examined to develop our final models are in Tables 1, 2, and 3. We collected demographic information (gender, race, age, income), substance use patterns (including type of substance used, frequency, common route of administration, and conditions or location of drug use), past medical history of comorbid conditions (hepatitis C virus [HCV], Methicillin-resistant Staphylococcus aureus [MRSA], any mental illness), measures of subsistence (access to food, shelter, clothing, and restrooms), chronic pain levels (Hawker et al., 2011), and poor sleep quality (Buysse et al., 1989). Regarding a history of Methicillin-resistant Staphylococcus aureus (MRSA) infections, we asked participants, “Has a health care professional ever told you that you have MRSA?” Given this self-reported measure, responses likely include MRSA infection and cases of MRSA colonization and should be interpreted broadly to include both.

Table 1:

Bivariate analysis of IDU-IE, SIRI specific symptoms and non-specific, untreated infection symptoms by selected socio-demographic characteristics among opioid-using people who inject drugs in Denver, CO and Los Angeles, CA, 2021/22 (n=472).

Characteristics Socio-demographics/socioeconomics	Total N=472 %	Non-specific infection symptoms, last 3 mos N=93 %	SIRI specific symptoms, last 3 mos N=64 %	Ever had infective endocarditis N=31 %
Study Site
Denver	53	57	64*	71*
Los Angeles	47	43	36	29
Gender
Male	77	70	2	52
Female	21	29	27	45*
Gender minority person	2	1	2	3
Race
White	52	53	56	52
Latinx	26	19	17	19
African American	6	5	8	3
Asian/Pacific Islander	1	3	2	3
Native American	10	12	12	16
Mixed Race	5	8	5	7
Age (years)
<30	16	18	20	7
30–39	35	33	31	42
40–49	24	26	27	29
≥50	25	23	22	23
Sexual orientation
Heterosexual male	66	57	58	52
Heterosexual female	14	24	19	39
~Gay/bisexual male	13	13	14	0
~Lesbian/bisexual female and/or GMP	6	6	9	10
High school education or more	77	77	83	81
Income per month (missing n=4)
< $1,000	53	58	53	61
$1,00 to $1,400	19	12	9	13
$1,401 to $2,100	13	15	19	16
$2,101 or more	15	15	19	10
Income source
Panhandling	35	39	45*	42
Unhoused in the last 3 months
Yes	84	88	91	77
Subsistence level (without shelter)
Low	34	24	17	35
Medium	31	29	30	23
High	35	47*	53*	42

^*p<0.05

^*Gender minority person (GMP): Participants who reported gender as transgender, gender non-conforming, and/or other

Table 2:

Bivariate analysis of IDU-IE, SIRI specific symptoms and non-specific, untreated infection symptoms and drug use characteristics among opioid-using people who inject drugs in Denver, CO and Los Angeles, CA, 2021/22 (n=472).

Characteristics Drug use characteristics	Total N=472 %	Non-specific infection symptoms, last 3 mos N=93 %	SIRI specific symptoms, last 3 mos N=64 %	Ever had infective endocarditis N=31 %
Drug use, last 3 months
Methamphetamine	84	82	84	87
Heroin	82	83	78	87
Fentanyl	70	82*	86*	77
Speedball~	29	26	25	42
Goofball~	63	63	63	77
Cannabis	75	68	66	68
Non-prescription use
Opioids	28	38*	36	36
Tranquilizers	38	46	47	39
Methadone	16	19	19	19
Buprenorphine	14	19	23*	19
Daily use, last 3 months
Speedball~	5	8	6	10
Goofball~	23	25	25	36
Methamphetamine	44	46	48	39
Heroin	42	40	34	61*
Fentanyl	34	36	38	52*
Cannabis	31	29	34	26
Non-prescription use
Opiates	6	3	5	10
Daily injection use
Goofball~	18	20	20	29
Heroin	39	39	34	61*
Fentanyl	10	16*	16	26*
Methamphetamine	17	20	22	10
Non-prescribed opioids	3	2	3	10*
Frequency of all opioid use+
Less than daily	24	25	28	3
1 to 2 times a day	30	26	30	29
3 or more times a day	46	49	42	68*
Frequency of injection, last 3 months
Less than daily	30	27	28	16
1 to 2 times a day	20	17	20	19
3 or more times a day	50	56	52	65
Places where injections occurred, last 3 months (missing n=1)
Own house/apartment	28	27	25	26
Friend’s house/apartment	48	46	50	65*
Abandoned building/garage/shed	33	40	48*	45
Vehicle/car	48	57	61*	65*
Shooting gallery	8	10	11	13
Motel/Hotel	47	53	58	58
Public setting	60	74	78*	65
Tent	62	58	61	81*
Public restroom	61	72	78*	65

^*p<0.05

⁺includes heroin, fentanyl, prescription opioids and combined use

^~goofballs [methamphetamine & heroin/fentanyl] or speedballs [cocaine & heroin/fentanyl])

Table 3:

Bivariate analysis of IDU-IE, SIRI specific symptoms and non-specific, untreated infection symptoms by selected health related, risk behaviors/protective factor among opioid-using people who inject drugs in Denver, CO and Los Angeles, CA, 2021/22 (n=472).

Characteristics Health related characteristics	Total N=472 %	Non-specific infection symptoms, last 3 mos N=93 %	SIRI specific symptoms, last 3 mos N=64 %	Ever hadinfective endocarditis N=31 %
Any mental health diagnosis	67	81*	80*	87*
Any Chronic Pain, last 3 months	55	69*	66	55
Poor quality sleep, last 3 month	62	74*	77*	58
Ever diagnosed with
Diabetes (missing n=2)	7	11	9	13
High BP/hypertension (missing n=2)	28	30	33	32
Congestive Heart Failure/Heart attack	6	7	6	13
Stroke (missing n=1)	5	3	3	13*
MRSA~ (missing n=1)	21	32*	33*	58*
HCV (missing n=1)	53	61	64*	90*
HIV	3	3	3	3
Withdrawal symptoms, last 3 months
Opioids	74	85*	88*	90*
Methamphetamines	41	52*	55*	55
Risk Behaviors/Protective factors
Injecting another PWID last 3 months
Yes	55	60	63	77*
Received an injection by PWID last 3 months
Yes	34	39	42	29
Drug equipment risk
Receptive syringe sharing	21	28	27	19
Cooker sharing	37	50*	55*	39
Rinse water sharing	24	31*	36*	32
Filter/cotton sharing	27	38*	44*	42*
Reused cotton (missing n=1)	72	73	72	84
Licked needle prior to injection	20	28*	34*	13
Used saliva to clean injection site	26	32	41*	39
Frequency of re-using cotton/filters
Never	29	27	28	16
Less than monthly	14	15	17	3
Monthly	21	15	13	26
Weekly	14	19	17	13
Daily	22	24	25	42*
Frequency of sharing cotton/filters
Never	74	62	56	58
Less than monthly	8	9	13	10
Monthly	11	20	9	23
Weekly	4	5	8	7
Daily	3	3	5*	3
Frequency of licking needle, last 3 months
Never	82	72	66	87
Less than half	11	9	25	13
More than half	7	9*	9*	0
Frequency of using saliva on injection site, last 3 months
Never	76	71	64	65
Less than half	19	24	27	32
More than half	5	5	9*	3
Rushed injection, last 3 months
None	36	27	20	39
1 to 9 times	29	29	30	10
10 times or more	35	44	50*	52*
Any syringe reuse, last 3 months	40	50*	52*	42
Any substance use treatment, last 3 months
Yes	26	29	27	26
Buprenorphine	13	16	19	16
Methadone mJntenance	20	23	22	26

^*p<0.05

^~Methicillin-resistant Staphylococcus aureus

Injection-related risk behaviors examined included receptive syringe sharing, injection equipment (cookers, cotton, rinse water) sharing, using salvia to wipe injection site before use, licking needles prior to injection, and reusing syringes and cotton/filters (Dahlman, Hakansson, et al., 2017) and frequency of these behaviors in the last 3 months. We also measured and assessed drug use patterns in the last 3 months including times injected, rushed injections, and giving and receiving injections from other PWID. Any diagnoses with HCV, MRSA, mental illness, and chronic pain was also considered potential risk factors of history of IDU-IE and current infection symptoms. We considered protective factors such as enrollment in substance use treatment, and type of treatment engagement in the last 3 months. Health variables examined included history of mental disorders, HCV/HIV, and common chronic conditions (such as pain, diabetes and hypertension).

Demographic and socioeconomic co-variates included race, gender, age, education, income in the last 3 months, income sources, and housing status. We also considered subsistence measures. Using Gelberg’s index, we asked participants about the difficulty obtaining food, clothing, toilets, and showers in the last 3 months (Gelberg et al., 1997). We summed responses to these items and divided the participants into thirds, from ‘low’ to ‘high’ (where a ‘high’ subsistence score meant greater material hardship).

2.3. Statistical Analyses:

The final multiple regression models were derived based on theoretical justification guided by extant literature and biological plausibility. We selected specific predictor variables from the following domains (sociodemographic, economic, health, risk behaviors and drug use) guided by existing studies and clinical significance. After descriptive analysis of variables under consideration, we conducted bivariate analyses to identify variables associated with each of our outcomes. We used chi-square tests for categorical variables and t-tests for continuous variables, with statistical significance at p<0.05. We progressed through five variable domains - sociodemographic, economic, health, risk behaviors, and drug use. Using the Pearson’s correlation coefficient, we assessed collinearity among variables within each domain. For a pair of collinear variables (rho > 0.30), variable less strongly associated with the outcome variable was excluded from the analysis. With this final set of independent variables, we constructed two multiple logistic regression models assessing association with 1) untreated, nonspecific infection symptoms and 2) SIRI symptoms. In a third multiple logistic regression model, we assessed independent variable association with IDU-IE diagnosis. Due to the small number of cases in our sample (n = 31), we constructed a model using the three most-associated independent variables under consideration. The remaining variables were entered into the model stepwise to determine if any remained significant. All analyses were conducted using SPSS, version 29.01 (IBM Corporation).

3. Results:

3.1. Sample Characteristics:

The sample (n=472) was predominantly male (77%) with an average age of 41 years (SD = 11.1; median = 39, Interquartile Range [IQR] = 32, 49). For the purposes of this analysis, we categorized individuals who identified as female, transgender, gender non-conforming or other together as ‘female and/or gender minority persons (GMP)’ (Table 1) because there were too few (less than 2% total) participants that reported being transgender, gender non-conforming and other to examine meaningfully as separate categories. Most participants were White (51%) followed by Latinx (26%), Native American (10%), Black/African American (6%), Asian or Pacific Islander (1%), and Mixed Race/Other Race (6%). Most participants reported being heterosexual (81%), followed by those who reported being gay, lesbian or bisexual (17%), and remaining reported being asexual (2%). Most participants (53%) made less than $1,000 per month from all sources of income (legal and otherwise). Most of the sample was current unhoused or had unstable housing (84%).

We assessed drug use in the previous three months. Participants reporting using methamphetamine (84%), heroin (82%), cannabis (75%), and fentanyl (70%). In terms of frequency of injection drug use, half the respondents injected drugs 3 or more times a day (50%) while the rest of the sample was reported injecting 1 to 2 times a day (20%) and less than daily (30%). Participants in the sample reported daily heroin injection (39%), followed by goofball (a combination of heroin and/or fentanyl with methamphetamine) (18%), methamphetamine (17%), and fentanyl (10%).

For our health outcomes, we examined ever having IDU-IE (7%), SIRI symptoms (14%), and untreated, non-specific infection symptoms (20%) in the last 3 months. Other health problems found in the sample included with any mental health diagnosis (67%), chronic pain (55%), ever having HCV infection (53%), ever having HAV infection (45%), ever having MRSA (21%), heart attack/heart failure (6%), stroke (5%), and reporting HIV (3%).

3.2. Factors Associated with ever having IDU-IE:

Bivariate analyses by outcome variables are presented in Table 1. In bivariate analysis of ever having IDU-IE, we found significant associations by demographic/socioeconomic characteristics (including but not limited to study site, gender), drug use patterns (daily heroin and fentanyl use; daily injection of heroin, fentanyl, and opioids; frequency of any opioid use), health conditions (mental illness, pain, HCV, stroke, and opioid withdrawal ), injection practices (shared filters/cotton, frequency of sharing filters/cotton, rushed injections, and injected another person) and places where participants injected (tents, abandoned buildings, friend’s house). In multiple logistic regression analysis (Table 2), we found the following variables to be independently associated with ever having IDU-IE: Ever being diagnosed with HCV infection (adjusted odds ratio [AOR]=8.37; 95% confidence interval [CI]=2.46, 28.49), history of MRSA infections (AOR=5.37; 95% CI=2.44, 11.80), and identifying as female and/or GMP (AOR=3.14; 95% CI=1.42, 6.95).

3.3. Factors associated with SIRI symptoms:

In bivariate analysis (Table 1), SIRI symptoms were associated with demographic/socioeconomic characteristics (study site, panhandling, subsistence index), drug use (fentanyl and buprenorphine use), health conditions (mental illness, poor sleep quality, history of MRSA, HCV, stroke, and opioid and methamphetamine withdrawal symptoms), injection practices (sharing filters/cotton, cookers, rinse water needle licking, wiping injection site with saliva, and frequency of filter/cotton sharing, needle licking, and wiping, along with syringe reuse and rushed injection) and places where participants injected (abandoned buildings, vehicle, public settings, and public restrooms). In multivariable logistic regression (Table 3), we found SIRI symptoms to be associated with high subsistence index level (AOR=2.47; 95% CI=1.17, 5.22 as compared to low subsistence), any fentanyl use in the last 3 months (AOR=2.15; 95% CI=1.01, 4.61), sharing filter/cotton in the last 3 months (AOR=1.93; 95% CI=1.10, 3.39) and licking needles prior to injection in the last 3 months (AOR=1.85; 95% CI=1.02, 3.36).

3.3. Factors Associated with untreated, non-specific infection symptoms:

In bivariate analysis of symptoms for untreated, non-specific infections (Table 1), we found demographic/socioeconomic (subsistence index level), drug use (fentanyl and opioids use, daily fentanyl use, and daily fentanyl injection), health conditions (mental illness, pain, poor sleep quality, history of MRSA, and opioid and methamphetamine withdrawal symptoms), and injection risk (sharing filters/cotton, water and cookers, syringe reuse, needle licking prior to injection, and frequency of needle licking). In multivariable logistic regression (Table 4), we found untreated, non-specific infection symptoms to be independently associated with poor quality sleep (AOR=2.04; 95% CI=1.21, 3.43), any mental health diagnoses (AOR=2.01, 95% CI=1.13, 3.56), any chronic pain (AOR=1.89; 95% CI=1.14, 3.11), sharing filter/cotton in the last 3 months (AOR=1.81, 95% CI=1.10, 2.98), and ever having MRSA (AOR=1.75; 95% CI=1.04, 2.97).

Table 4:

Multivariate logistic regression of ever being diagnosed with IDU-IE among PWID in Denver, CO and Los Angeles, CA 2021/22 (n=470)

	Adjusted Odds Ratio (AOR) 95% confidence interval	P=
Ever diagnosed with HCV
No	Referent
Yes	8.37 (2.46, 28.49)	<0.001
Ever diagnosed with MRSA~
No	Referent
Yes	5.37 (2.44, 11.80)	<0.001
Female and GMP*
No	Referent
Yes	3.14 (1.42, 6.95)	<0.001

^~Methicillin-resistant Staphylococcus aureus;

^*Includes times used all opioids (i.e., heroin, fentanyl, opioid prescription medications, goofballs [heroin/fentanyl with methamphetamine] and speedball [heroin/fentanyl with cocaine])

^*Gender minority person (GMP): Participants who reported gender as transgender, gender non-conforming, and/or other

4. Discussion:

To our best knowledge, this is the first study to examine IDU-IE infection symptomology in a community sample. For participants who had a history of IDU-IE, prior infection with HCV and history of MRSA was more likely. Extant literature has found HCV infection is high (36–82%) among IDU-IE patients (Schranz & Barocas, 2020). Further, HCV and IDU-IE have some overlapping risk factors (i.e., syringe sharing, injection equipment sharing, and other practices) that facilitate bacterial and viral contamination (Goldshear et al., 2021) for infection and reinfection. To address this, there is a need for comprehensive, evidence-based substance use disorder treatment for patients with HCV - this includes pharmacotherapy, multi-disciplinary teams (Cooper, 2008; Fadnes et al., 2021; Olea et al., 2018), and low-barrier harm reduction based healthcare (Milne et al., 2015; Rizk et al., 2019). The current HCV treatments have become simpler to use and have shorter durations to achieve HCV suppression; making these treatments available to PWID with HCV should be a priority (Bruggmann & Grebely, 2015; Falade-Nwulia et al., 2019; Grebely et al., 2015; Zeremski et al., 2013).

MRSA infections are significantly more common among patients with IDU-IE compared with patients with non-IDU-IE (Damlin & Westling, 2021). A systematic review examining temporal changes in infective endocarditis epidemiology found Staphylococcus aureus to be the most common causative pathogen (Talha et al., 2021). Emerging literature links community acquired MRSA infections to IDU-IE, risk factors for which include injection drug use, criminal-legal involvement, sanitation and hygiene issues (Millar et al., 2008). In our study, 53% of PWID made less than $1,000 per month and 84% reported current housing insecurity and/or instability indicating pervasive structural vulnerabilities (Leibler et al., 2017; Leibler et al., 2019; Millar et al., 2008; Packer et al., 2019). Previous research suggests that partial oral antibiotic therapy for PWID in complicated Staphylococcus aureus (S. aureus) bloodstream infections, including infective endocarditis, is significantly protective against death and microbiologic failure (Wildenthal et al., 2023). Moreover, patients who received oral antibiotics despite an incomplete intravenous antibiotic course were significantly less likely to die or experience microbiologic failure than patients discharged without oral antibiotics (Wildenthal et al., 2023). Considering the high rate of self-discharges or patient directed Discharges (PDD) among PWID (Appa et al., 2020; Ti et al., 2015), developing discharge protocols accordingly, is a critical intervention point (Sharma et al., 2017). Clinical harm reduction strategies for unhoused and indigent PWID who may self-discharge before completing anti-biotic treatment should be prioritized to reduce MRSA bloodstream infections by optimizing intervention in acute care and ED settings (Bourgois et al., 2017; Parikh et al., 2020). Notably, patients with SIRI who self-discharged show high rates of antibiotic adherence further emphasizing the necessity and potential effectiveness of these measures (Lewis et al., 2022).

In addition, we found that identifying as female and/or GMP was associated with ever having IDU-IE. Some but not all studies of patients have found higher proportion of women have IDU-IE or report have poorer clinical outcomes (Kimmel et al., 2020; Shah et al., 2020; Wright et al., 2018; Wurcel et al., 2016) and women have long had higher odds of skin and soft tissue infections (Baltes et al., 2020; Dahlman et al., 2015; Larney et al., 2017; Lloyd-Smith et al., 2008). Studies examining sex-specific in IDU-IE have yet to determine causality (McCrary et al., 2024; Bhandari et al., 2022). Interestingly, Bhandari and colleagues reported that female sex was associated younger age, higher rates of self-discharge, and substance use (Bhandari er al., 2022). Improving access to harm reduction services and addressing gaps in clinical care for female and GMP is warranted. Several factors within the intersectional risk environment put female and gendered persons at higher risk for drug related harms (Collins et al., 2019). These include risk of intimate partner violence (Frye et al., 2007; Simmons et al., 2015; Hargrave et al., 2024). Additionally, women are more likely to be injected by someone else, and trade drugs for sex (Lorvick et al., 2006; Shannon et al., 2008; Sherman et al., 2001; Spittal et al., 2003; Strathdee et al., 2008). Female sex workers who trade sex are also more likely to have a partner who injects drugs and obtain syringes from street sellers and be injected by another person (Park et al., 2019). Overlapping vulnerabilities in the risk environment contribute to increased likelihood of serious sequelae from injection drug use among women (Astemborski et al., 1994; Collins et al., 2019; Nerlander et al., 2017; Wurcel et al., 2018). Women, in particular, have historically had poorer access to treatments (Gao et al., 2023; Rosenbaum & Murphy, 1981), particularly in the context of pregnancies and child rearing responsibilities. Women with substance use disorders and IE may also experience more stigmatization than their male counterparts while in care. Our findings underscore a need for tailored interventions and female and GMP specific resources and education.

Two behaviors that are subject to intervention were associated with SIRI symptoms – sharing cotton or filters and licking needles prior to injection as has been found in at least one case-control study (Shah et al., 2020). Both are highly plausible modes of bacterial infection and have been associated with skin and soft tissue infections (Allaw et al., 2023; Dahlman, Hakansson, et al., 2017; Marks et al., 2024). They can be addressed with improved access to safer injection supplies, safe consumption sites, and educational efforts to reinforce the risk associated with needle-licking prior to injection (Hochstatter et al., 2020; Lalanne et al., 2024). Harm reduction education for bacterial infections include disseminating information about safe injection practices, address risks of injection behaviors such as needle licking, and signs and symptoms of serious injection related infections (Bahji et al., 2020; Peckham & Young, 2020; Sharma et al., 2017). Further education should be developed based on qualitative data from structurally vulnerable PWID to ensure efficacy and uptake (Phillips et al., 2013). Testing of educational efforts and implementation of proven interventions should be pursued to reduce SIRI symptoms in this population. Harm reduction interventions for health education related bacterial infections should be developed based on qualitative data from PWID to ensure efficacy and uptake (Phillips et al., 2013). The following should be considered 1) delivering interventions via established community partners (Wegner et al., 2024) 2) expanding access to safer smoking supplies to that may reduce risks associated with injection routes of administration (Kral et al, 2021), 3) evidence-based measures such as safe consumption sites (Lambdin et al., 2022; Foreman-Mackey et al., 2019), and 4) low barrier access to medications for opioid use disorder (Barocas et al., 2021b).

Fentanyl use was associated with SIRI symptoms. Up until very recently, ‘black tar’ heroin was the predominate opioid in Los Angeles, CA and Denver, CO. Such heroin has long been associated with increased risk for toxins (botulism) and bacterial agents (including tetanus and Clostridium sordellii) (Hoffman & Yee, 2023; Kimura et al., 2004; Larney et al., 2017; Summers et al., 2017). Whether this association persists following the rapid change to smoking fentanyl in Western states is an open question (Ciccarone et al., 2024; Eger et al., 2024; Kral et al., 2021; O’Donnell et al., 2021; Tanz et al., 2024). While most bacterial infections among people who use drugs are associated with injection drug use, opioid induced immunosuppression may be a possible risk factor for systemic infections that would not necessarily require injection use (Bettinger & Friedman, 2024; Sun et al., 2023; Wei et al., 2003). Preliminary analyses by our investigators suggest that growing fentanyl smoking is associated with lower odds of self-reported abscesses (Bluthenthal et al., 2023). Studies examining the association between transitions in mode of use and bacterial infections seems warranted. More investigation is needed to understand the potential benefits of smoking to prevent soft tissue infections versus the contamination risk from bacterial agents in fentanyl, as well as opioid induced immunosuppression.

High material deprivation (i.e., difficulty accessing basic needs such as clothing, food, showers, and restrooms) was associated with SIRI symptoms. Difficulty accessing subsistence needs such as clothing can indicate overlapping structural vulnerabilities. A Water, Sanitation, and Hygiene (WaSH) insecurity study among unhoused participants on Skid Row in Los Angeles reported that access to WaSH services is most challenging at night, with 91% of people washing their items less than 3 times a month due to laundry inaccessibility (Avelar Portillo et al., 2023). Participants in their study found it easier to throw away clothing than maintain items or laundered items in sinks, businesses or public bathrooms due to the barriers to accessing WaSH services (Avelar Portillo et al., 2023). Barriers include long wait times, inconvenient hours, or out-of-service facilities at the public restrooms in parks and libraries and restrooms from non-profit organizations (e.g., shelters, soup kitchens, mobile showers, and religious organizations) that they used to bathe, wash clothes, or use the washroom (Avelar Portillo et al., 2023). WaSH services are critical to reducing the risk of infection; access to potable drinking water, showers, antibacterial hand-soaps, and laundry services needs urgent expansion to 24 hour services as well as multiple centers, as recommended by authors in the WaSH study (Avelar Portillo et al., 2023). This is likely and especially challenging for people who menstruate and do not have access to clothing and/or WaSH services - while there is limited data in this area, the WaSH study reports that participants in their study use toilet paper and clothing items to manage menstruation, emphasizing the need to expand access to menstrual products and WaSH services (Avelar Portillo et al., 2023).

Along with sharing cotton/filters and history of MRSA infection (as discussed above), untreated and non-specific infection symptoms were also associated with chronic pain, any mental health diagnosis, and poor quality sleep. Existing literature has found that mental illness is associated with higher morbidity from polysubstance use (Barocas et al., 2019). Given the high rate of diagnosed mental illness, the gap in mental health care services for this population is evident as are the consequences of these gaps. The need for acute mental healthcare services, long-term follow up care, integrated harm reduction based mental health services (Pettit Bruns & Kraguljac, 2023), street-based mental health care, and expanded access to medications for psychiatric diagnoses is clear and urgent. Currently, there is a lack of comprehensive clinical guidelines for co-occurrences of opioid use disorder (OUD) and serious mental illness, the development of these guidelines must be based on harm reduction and considering the structural determinants of mental illness and substance use (Pettit Bruns & Kraguljac, 2023).

Chronic pain, among people who use drugs, lies at the intersection of three overlapping factors 1) early childhood experiences, 2) hyperalgesia (increased sensitivity to pain), and structural vulnerabilities. Existing data shows that individuals with adverse and traumatic early developmental experiences are more likely to have chronic pain (Davis et al., 2005) and increased sensitivity to pain (Caes & Roche, 2020; Tesarz et al., 2016) and those experiences are also risk factors for developing OUD (Hser et al., 2017; Nazarian et al., 2021; Orhurhu et al., 2019; Speed et al., 2018). Beyond that, increased tolerance and hyperalgesia from the OUD itself can contribute to pain experienced by people who use drugs (Lee et al., 2011). When considering the above in the context of major structural vulnerabilities such as housing insecurity and instability, extant data shows pain is more prevalent among people who are unhoused (Fisher et al., 2013; Hwang et al., 2011; Matter et al., 2009). Integrating the above, it is very likely that chronic pain plays a bidirectional role with OUDs, especially with 43.2% of PWID reporting chronic pain as a primary factor in their opioid re-initiation opioid use (Ellis et al., 2021). More data is required to understand how chronic pain is related to bacterial infection symptomatology, especially as they are associated with structural vulnerabilities such as housing and access to healthcare. There is an urgent need for the development of harm reduction based pain management protocols specific to chronic pain patients with OUD (Kakko et al., 2018) - these protocols must be highly patient-centered and consider factors such as hyperalgesia among OUD patients. This includes patients who have been on methadone for extended periods of time as data suggests that standard doses of morphine may be ineffective for pain management in these patients (Doverty et al., 2001). Previous literature shows that chronic pain is prevalent among opioid-using PWID (Dahlman, Kral, et al., 2017), and changing/increasing injection frequency and dosage to manage acute episodes of pain (R. E. Harris et al., 2018). Chronic pain could also be indicative of existing and unmanaged infection symptoms (O’Donnell & Lawson, 2016). Future qualitative research can explain the role of chronic pain management, injection frequency, and patient experiences in healthcare settings. A report on intensive care as an overdose management tool outlines a framework of Directly-Observed Therapy for Pain (DOT-P) – DOT-P consists of specialty intensive pain care clinics where medical professionals provide pharmaceutical-grade opioids (for instance, hydromorphone) in a monitored-settings in addition to resources and services that are address the needs of highly marginalized patients with OUD and co-occurrence of chronic pain (Anderson et al., 2019).

The social and environmental determinants of sleep on the health of PWID should be an important future focus for research, particularly among unhoused PWID. Our study found that a factor associated with untreated infection symptoms was poor sleep quality. Sleep quality as a social determinant of health can be associated with the risk environment of the individual (Moore et al., 2023; Rhodes et al., 2003). Qualitative studies have shown how type of drug use is mediated by sleep due to concerns about safety and protection of belongings among unhoused people who use drugs (Mayock et al., 2015; McKenna, 2013; Milliken, 2023). Sleep also plays a critical role in modulating immune responses to infectious diseases in cases of SIRIs and IDU-IE (Irwin, 2012). Existing research also reports gender differences in sleep quality among people who use drugs (He et al., 2020) - with female methamphetamine users reporting the worst sleep quality - more qualitative research is required to contextualize these findings given structural drivers such as safety, protection of belongings, fear of violence shaping stimulant-use among female identifying people who use drugs (Jones et al., 2023; Lorvick et al., 2012; McKenna, 2013; Riley et al., 2015; Stahlman et al., 2013).

4.1. Limitations:

Study limitations include self-reported data, community-recruitment, and cross-sectional study design. This means that all our outcomes were generated from patient reported diagnoses. In this study, we intended to examine as patient perspectives from a community sample as that is underrepresented in the existing body of literature. Future research should assess concordance between symptomatology among people who inject drugs with a history of endocarditis. Self-report data are subject to recall and social desirability bias. However, in reliability and validity studies among PWID reports have been found to have acceptable psychometric properties (Dowling-Guyer et al., 1994; Weatherby et al., 1994). In this study, Participants were asked “Has a health care professional ever told you that you have MRSA?” This is a self-reported measure that has limitations. MRSA infections can be interpreted broadly. For example, some patients may be told that they are MRSA positive when they are in the hospital, but this may be because they are colonized with the bacteria and may not necessarily have an infection. Additionally, some people may state that they had MRSA when it was staph aureus. That said, it is important to also note that MRSA colonization is independently associated wounds and skin and soft tissue infections among PWID which when untreated can lead to endocarditis (Sanchez et al., 2021). Additionally, this study did not use chart review to verify patient medical history, all medical history was self-reported and while the survey was comprehensive in its examination of medical history, it is possible that participants may omit, misunderstand, or miscommunicate their medical history. Further, the baseline survey did not explicitly include items pertaining to xylazine however, we did ask about sedatives and are reporting results accordingly. This was because initial reports of xylazine appeared after baseline data were collected.

There are a few reasons for variation between factors associated with ever having IDU-IE, SIRI symptoms, and non-symptoms. We asked participants “Have you ever been diagnosed with infective endocarditis (an infection to your heart valve)?”– however, we do not know when the diagnosis was made. Associated factors in the model included ever having MRSA, ever having HCV, and being female. On the other hand, the items about non-specific infection symptoms and SIRI symptoms were specifically about the past 3 months. Accordingly, the associated factors may differ based on occurrences in the last 3 months vs life history. These also provide important insight into factors associated recently vs lifetime, and we have retained both for purposes of identifying behaviors that might prevent endocarditis and other serious infections in this high-risk and structurally vulnerable population.

Participants were recruited from syringe service programs (in Los Angeles and Denver) and community sites frequented by PWID such as parks and shower facilities (in Los Angeles). This approach was taken since data collection occurred during the COVID-19 pandemic and typical approaches that rely upon snowball sampling such as respondent driven sampling and target sampling (Heckathorn, 1997; Watters & Biernacki, 1989), were not feasible. Therefore, data collection is not representative of PWID in Los Angeles or Denver. The study design was cross sectional and future research in this area should evaluate whether the associations from this study continue to remain consistent in longitudinal studies.

4.2. Conclusions:

Improvements in biomedical treatments for opioid use disorder, mental health, pain, and HCV are impressive and should be made widely accessible to PWID given their elevated risk. Hospitals, substance use treatment centers, and SSPs should increase efforts to provide PWID with sufficient amounts of injection supplies to reduce sharing and/or reuse of injection materials. Needle licking is uncommon (less than 20%) but appears to hold significant risk. Educational and behavioral interventions to reduce this behavior are indicated. Lastly, basic needs among PWID appear to be rarely met; if we want to improve health in this population, we need interventions that address alleviate material hardship by addressing aspects of well-being such as food, shelter, clothing, and personal hygiene.

Table 5:

Multivariate logistic regression model of SIRI specific infection symptoms among PWID in Denver, CO and Los Angeles, CA 2021/22 (n=472)

	Adjusted Odds Ratio (AOR) 95% confidence interval	P=
Subsistence index
Low	Referent
Medium	1.55 (0.69, 3.46)	0.29
High	2.47 (1.17, 5.22)	0.02
Any fentanyl use, last 3 months
No	Referent
Yes	2.15 (1.01, 4.61)	0.05
Shared filter/cotton, last 3 months
No	Referent
Yes	1.93 (1.10, 3.39)	0.02
Licked needle prior to injection, last 3 months
No	Referent
Yes	1.85 (1.02, 3.36)	0.04

Table 6:

Multivariate logistic regression model of non-specific infection symptoms among PWID in Denver, CO and Los Angeles, CA 2021/22 (n=471)

	Adjusted Odds Ratio (AOR) 95% confidence interval	P=
Poor quality sleep
No	Referent
Yes	2.04 (1.21, 3.43)	0.0008
Any mental health diagnoses
No	Referent
Yes	2.01 (1.13, 3.56)	0.02
Any chronic pain
No	Referent
Yes	1.89 (1.14, 3.11)	0.01
Shared filter/cotton, last 3 months
No	Referent
Yes	1.81 (1.10, 2.98)	0.02
Ever diagnosed with MRSA~
No	Referent
Yes	1.75 (1.04, 2.97)	0.04

^~Methicillin-resistant Staphylococcus aureus

Highlights:

• Infective endocarditis was associated with Hepatitis C and prior MRSA infection.

• Infective endocarditis was associated with identifying as female and/or GMP.

• Serious injection related infection (SIRI) symptoms were associated fentanyl use.

• SIRI symptoms were associated with equipment sharing and needle licking.

• SIRI symptoms were associated with greater material hardship.

• Findings emphasize the need for structural harm reduction interventions.