Influence of Social Determinants of Health and Substance Use Characteristics on Persons who Use Drugs Pursuit of Care for Hepatitis C Virus Infection

Abstract

Background:

Social determinants of health (SDOH), i.e., social, behavioral and environmental factors, are increasingly recognized for their important influence on health outcomes. Data are limited on the influence of SDOH, substance use characteristics, and their interactions on pursuit of hepatitis C virus (HCV) care among individuals with opioid use disorder (OUD). Linkage to HCV care remains low in this population despite high HCV prevalence and incidence.

Aims:

To investigate the influence of SDOH, substance use factors, and their interactions on HCV treatment uptake among OUD patients in a methadone treatment program.

Methods:

Information on patient demographics, SDOH, substance use characteristics, and co-morbid medical conditions were obtained from the paper and electronic medical records of OUD patients on methadone. We applied multiple correspondence analysis, k-means algorithm, and logistic regression with least absolute shrinkage and selection operator penalty to identify variables and clusters associated with pursuit of HCV care.

Results:

Data from 161 patients (57% male, 60% Caucasian, mean age 45 years) were evaluated. Being employed, the absence of support systems, and a history of foster care were the strongest positive predictors of treatment pursuit. The use of crack/cocaine as the initial illicit substance, criminal activity without incarceration, and the absence of a family history of chemical dependency were the strongest negative predictors. We identified clusters among persons with OUD based upon their likelihood to pursue HCV management.

Conclusion:

Utilizing data from the medical record, we were able to identify factors positively and negatively associated with linkage-to-care for HCV. We were also able to divide patients into clusters of factors associated with linkage-to-care for HCV. These results could be used to identify individuals with OUD based upon their readiness for HCV care.

1. Introduction

Social determinants of health (SDOH) are social, behavioral, and environmental factors that contribute to health inequalities and play a disproportionately large role in health outcomes. As SDOH contribute substantially to an individual’s overall physical and mental health, they can have a deleterious effect on health outcomes (Daniel, Bornstein, & Kane, 2018). For example, low education levels, racial segregation, poverty, low income, and low social support contribute substantially to mortality (Galea, Tracy, Hoggatt, DiMaggio, & Karpati, 2011). To improve health outcomes, attempting to moderate SDOH should be prioritized (Heiman & Artiga, 2015; Taylor et al, 2016). Recently, professional societies, such as the American College of Physicians, have highlighted research gaps in the area of SDOH, particularly with regard to the inclusion of disadvantaged and underserved populations (Daniel et al., 2018; Thomas-Henkel & Schulman, 2017).

SDOH disproportionally affect vulnerable populations, such as persons with substance use disorders (PWSUD) (Galea & Vlahov, 2002). In PWSUD, social factors affect both substance use behaviors and the health consequences of use (Galea & Vlahov, 2002; Wilkinson & Marmot, 2003). While SDOH have been widely investigated in HIV and in sexually transmitted infections, data are limited on their relationship to adverse health outcomes in hepatitis C virus (HCV) infection. HCV is a common blood born infection that can lead to cirrhosis and to liver cancer and that disproportionately affects those with lower socioeconomic status (Bruden et al, 2017; Ditah et al, 2014). As the leading HCV transmission route is injection drug use, PWSUD constitute the vast majority of incident (75%) and prevalent (80%) HCV cases in developed countries including the United States (Alavi et al, 2014; Ditah et al., 2014; Hajarizadeh, Grebely, & Dore, 2013; Shepard, Finelli, & Alter, 2005; Zibbell et al, 2018). From 2006 to 2012, the incidence of acute HCV increased 364% among those less than 30 years of age in the areas of the United States most affected by the opioid epidemic (Zibbell et al, 2015). Despite the frequency and recent increases in HCV incidence among PWSUD, this population remains difficult-to-engage in HCV care.

The changes in HCV demographics in the United States underscore the importance of elucidation of the influence of SDOH, substance use variables, and their interactions on linkage-to-care for the viral infection. Identifying the most significant obstacles toward PWSUD’s pursuit of HCV care could foster individualized treatment approaches. In this work, we evaluated the influence of social and substance use factors as well as their interactions on linkage-to-care for HCV. We conducted a retrospective review of the medical record at a methadone treatment program (MTP) to discern SDOH and substance use factors positively and negatively associated with linkage-to-care for HCV. We also assessed the interactions between individual factors to identify clusters of individuals that could be targeted to improve PWSUD participation in HCV care.

2. Methods

2.1. Participants and procedure

In this retrospective cross-sectional study, we reviewed the paper charts and electronic medical records (EMR) of all patients currently enrolled at an MTP in the greater Western New York (WNY) area. The study protocol was approved by the University at Buffalo Institutional Review Board. The community-based MTP had 481 active patients and had a turnover rate of approximately 10 patients per month at the time of data collection. The demographics of the MTP clients were similar to those at three other MTPs in the WNY region (Supplementary Table 1).

The MTP employs administrators, nurses, and counselors who administer methadone to individuals with opioid use disorder (OUD). Upon enrollment, patients are assisted by a counselor to complete a comprehensive intake form that includes information mandated by New York State Office of Alcohol and Substance Abuse Services as well as additional demographic, psychosocial, and reflective responses (e.g., client’s perception of problems in medical, mental health, and social domains as well as items assessing self-awareness). Two researchers (NW and NB) reviewed the medical record and collaborated to ensure consistency in data extraction. Data were extracted from the past medical history, past or current laboratory reports, and/or the practicing physician’s “impression” sections of the medical record. If any of the three areas were incomplete for a patient’s EMR entry, then the patient’s paper chart was reviewed. The two researchers conversed on a daily basis on-site at the MTP to avoid any discrepancies between medical chart language, laboratory results, and the information collected. The researchers also compared the process of data recording to ensure consistency.

Patients’ records were reviewed if they: 1) had an admission date to the MTP preceding January 1^st, 2017 (within 6 months of the records review) and 2) had a history of a positive HCV antibody test. The final sample consisted of 161 patients who fit both inclusion criteria.

2.2. Statistical analysis

We performed statistical analysis using R software (http://www.r-project.org/). We summarized categorical variables as counts and/or percentages, and continuous variables by their means and standard deviations, as appropriate. The data consist of 161 observations, one outcome variable (HCV linkage-to-care status) and 17 explanatory variables or covariates that were assigned into four different categories: demographic, SDOH, substance use, and co-morbid medical conditions. Specifically, four variables were evaluated under the heading of “demographic” variables, eight categorical variables were evaluated under the category “SDOH” variables, four variables under “substance use”, and we also evaluated the presence of medical co-morbidities.

We were interested in how strongly and in which way these variables are interrelated, and based on these associations, we attempted to group all patients into different clusters. This reduced the sample size from 161 to 155 individuals, with an approximate percentage of missingness of 3.7%. We first excluded individuals with missing demographic variable(s) from the analysis, and this reduced the sample size from 161 to 155. Missing demographic data, such as ethnicity and primary language from both the EMR and the paper chart, resulted in the exclusion of six individuals. Under the missing-at-random assumption, the missing values in the categories of SDOH and substance use were imputed using the R package “missMDA” (Josse & Husson, 2016). The percentage of missing data for each variable is illustrated (Supplementary Table 2). The variables included in the analysis, stratified by linkage-to-care status, are also illustrated (Table 1). We next utilized multiple correspondence analysis (MCA) via the R package “FactoMineR” to reveal patterns of relationships of SDOH and substance use variables as performed previously (Lê, Josse, & Husson, 2008; Talal et al, 2018). The analysis consists of three steps as described below. The first step consists of performing an MCA on SDOH and substance use variables. The purpose of this step is to explore the relationships between the different levels of social and substance use variables, all measured on a nominal or categorical scale. The two sets of variables, SDOH and substance use, were highly correlated and MCA is able to reveal the relationship hidden among them. Application of MCA to the imputed SDOH and substance use variables produced 18 principal dimensions (Supplementary Table 3). To decide the number of principal dimensions used for the purpose of exploring the relationships of these variables, we followed the recommendations by Gifi and retained the first two principal dimensions (Gifi, 1990). These two principal dimensions explain 27.6% of the total variance.

Table 1:

Variable and Participant Overview

Category	Variable	Level	All (n=161)			Evaluated (n=33)			Not Evaluated (n=128)
			Size	Mean/#	SD/%	Size	Mean/#	SD/%	Size	Mean/#	SD/%
	Age		161	45.2	13.4	33	48.2	13.4	128	44.27	13.3
	Gender	Female	161	70	43.5	33	8	24.2	128	62	48.4
		Male		91	56.5		25	75.8		66	51.6
		African American		25	15.7		8	24.2		17	13.5
Demographic	Race	Caucasian Hispanic/Native American	159	96 38	60.4 23.9	33	17 8	51.5 24.2	126	79 30	62.7 23.8
	Primary language	English Spanish	157	133 24	84.7 15.3	31	27 4	87.1 12.9	126	106 20	84.1 15.9
		Homeless/shelter/othersa		41	27.5		7	25.9		34	27.9
	Living situation	Lives alone	149	23	15.4	27	2	7.4	122	21	17.2
		Living with someone		85	57.1		18	66.7		67	54.9
		Divorced / othersa		48	30.2		10	31.3		38	29.9
	Marital status	Married	159	24	15.1	32	8	25.0	127	16	12.6
		Single		87	54.7		14	43.8		73	57.5
		Employed		18	11.2		7	21.2		11	8.6
	Employment status		161			33			128
		Unemployed		143	88.8		26	78.8		117	91.4
		GED		23	14.4		4	12.5		19	14.8
	Highest educational	High school(+VOC)		69	43.1		12	37.5		57	44.5
			160			32			128
	level	Higher education		52	32.5		12	37.5		40	31.3
		Less than high school		16	10.0		4	12.5		12	9.4
Social determinants of health		Criminal activity without incarceration		59	37.3		8	26.7		51	39.8
	Forensic history		158			30			128
		Incarceration		58	36.7		11	36.7		47	36.7
		No		41	26.0		11	36.7		30	23.4
		No		27	18.0		9	30.0		18	15.0
	Support systems		150			30			120
		Yes		123	82.0		21	70.0		102	85.0
	Victim of sexual	No		101	70.1		21	77.8		80	68.4
			144			27			117
	assault	Yes		43	29.9		6	22.2		37	31.6
		No		132	94.3		25	96.2		107	93.9
	Foster care history		140			26			114
		Yes		8	5.7		1	3.9		7	6.1
		<=16		85	57.4		20	69.0		65	54.6
	Age of initial use		148			29			119
		>16		63	42.6		9	31.0		54	45.4
	Family history of	No		55	38.2		12	41.4		43	37.4
			144			29			115
	chemical dependency	Yes		89	61.8		17	58.6		72	62.6
Substance use		Alcohol and/or othersb		42	28.2		9	31.0		33	27.5
	Substance of initial	Crack/cocaine		15	10.1		1	3.5		14	11.7
			149			29			120
	use	Heroin/opiates and/or
				92	61.7		19	65.5		73	60.8
		othersc
		Heroin/cocaine/opiate
				78	48.5		18	54.6		60	46.9
	Other illicit substance	and othersa	161			33			128
		Only		83	51.6		15	45.5		68	53.1
		heroin/cocaine/opiate
		Cardiovascular		22	13.7		5	15.2		17	13.3
Medical status	Comorbidity	Cardiovascular and othersd	161	32	19.9 33		9	27.3 128		23	18.0
		None		47	29.2		7	21.2		40	31.3
		Otherse		60	37.3		12	36.4		48	37.5

^aNo additional information available to explain the meaning of “others”.

^bHeroin, Cocaine, Xanax, and other substances. No additional information available to explain the meaning of “others”.

^cOthers: Crack/cocaine

^dOthers: Endocrine, musculoskeletal, respiratory, neuromuscular, gastrointestinal, immune, renal, neurologic, and others. No additional information available to explain the meaning of “others”.

^eOthers: Endocrine, musculoskeletal, respiratory, neuromuscular, gastrointestinal, immune, renal, urologic, ophthalmologic, neurologic, and others. No additional information available to explain the meaning of “others”.

Abbreviations: Standard deviation, SD; general educational development, GED; vocational training, VOC

The second step builds upon the first, as it uses the principal dimensions, produced by MCA, to further explore the relationships between SDOH and substance use variables. The implementation of K-means algorithm on the first two principal dimensions (i.e. dimensions explain the variability of a data set) of SDOH and substance use variables grouped all patients into three different clusters. We next used the elbow plot method to determine the number of clusters; it consists of plotting the within sum of squares as a function of different numbers of clusters k (Supplementary Fig. 1). The identification of a bend in the plot indicates the appropriate number of clusters.

The third and last step uses logistic regression with least absolute shrinkage and selection operator (LASSO) penalty to identify factors that contribute to an individual’s pursuit of HCV management. The covariates were demographic, co-morbid medical condition variables, and all principal dimensions of the SDOH and substance use variables. Our decision to include all principal dimensions was based on Jolliffe (Jolliffe, 1982). The use of penalized logistic regression, as opposed to standard logistic regression, is necessary because the number of observations per variable after representing all k-level categorical variables with k-1 variables and adding all principal dimensions in the model, is very low (approximately 3 observations per variable). In this case, standard logistic regression produces biased and unstable results. Thus, logistic regression with LASSO was applied using the R package “glmnet” (Park & Hastie, 2007). The outcome variable was whether or not an individual linked to care for HCV: 1 if the individual was evaluated for HCV and 0 if not. To select the tuning parameter, we used 5-fold cross-validation, producing a value of 0.052.

3. Results

3.1. Socio-demographic

Data were evaluated from a total of 161 individuals, and their characteristics are illustrated in Table 1. The average age of the individuals was 45 + 13.4 years, and 57% were male. The majority were Caucasian (60%), single (55%), and had a high school or general educational development (54%) diploma. Eighty-nine percent of individuals were unemployed, 57% lived with at least one other person, and 62% had a family history of chemical dependency. A total of 34% of individuals reported heroin, opiates, and/or other substances as the initial substance used with 57% of individuals reporting the age of initial use as ≤16 years, and 26% reporting no history of criminal activity.

Only 29% of the individuals did not have a medical co-morbidity.

3.2. Participant identification by cluster of factors associated with linkage-to-care for HCV

A plot of the number of clusters versus within-cluster sum of squares identifies that the appropriate number of clusters is 3 (Supplementary Fig. 1). Categories of SDOH and substance use variables are also plotted, and category labels within each cluster present the main characteristics of the individuals in that cluster. Table 2 presents the characteristics of individuals in different clusters, and Figure 1 displays the characteristics defining each cluster. We note that variables that belong to the same cluster are interrelated. Specifically, individuals in Cluster 1 tend to have an unstable living situation, are more likely to be divorced, unemployed, less educated, report having no support systems available, and started using illicit substances before the age of 16 years. In contrast, individuals in Cluster 3 are more likely to be married or single, are living with someone, are employed and have at least a high school education and reported the first illicit substance use after the age of 16 years. Cluster 2 individuals are characterized by having a history of sexual assault, a history of foster care and reporting their first illicit substance use as cocaine/crack (as opposed to alcohol and heroin/opiates in clusters 1 and 3, respectively).

Table 2.

Characteristics of patients in different clusters.

Category	Variable	Cluster 1 (n=68)	Cluster 2 (n=12)	Cluster 3 (n=75)
	Living situation	Homeless/shelter/others or lives alone		Living with someone
	Marital status	Divorced / others		Married or single
	Employment status	Unemployed		Employed
Social determinants of health	Highest educational level	Less than high school or high school(+VOC)		GED or higher education
	Forensic history		No	Criminal activity without incarceration or incarceration
	Support systems	No		Yes
	Victim of sexual assault	No	Yes
	Foster care history	No	Yes
Substance use	Age of initial use	≤ 16		> 16
	Family history of chemical dependency	No		Yes
	Substance of initial use	Alcohol and/or others	Crack/cocaine	Heroin/opiates and/or others
		Only		Heroin/cocaine/opiate and
	Other illicit substance
		heroin/cocaine/opiate		others

Abbreviations: General educational development, GED; vocational training, VOC

Fig. 1.

Plot of individuals and categories of social determinant and substance use variables. Individuals are plotted using different colors and shapes according to the clusters obtained by the K-means algorithm applied on the first two principal dimensions of social determinant and substance use variables.

3.3. Factors affecting linkage-to-care status

Only gender and the sixth principal dimension of the SDOH and substance use variables were selected as significant by the logistic regression with LASSO technique. The coefficient of the sixth principal dimension is 1.546 (p-value = 0.021) and the gender coefficient is 1.015 (p-value = 0.026). This translates into the following observation: an increase in the value of the sixth principal dimension of the SDOH and substance use variables could improve the probability of undergoing an HCV evaluation with relative risk = 4.69; similarly, male gender has a higher probability of undergoing an HCV evaluation compared to female gender with relative risk = 2.76.

Further analysis was conducted to determine if certain variables were positively or negatively correlated with treatment uptake. Figure 2 presents the coordinates of categories of variables on the sixth principal dimension of SDOH and substance use variables. In this figure, any category with a positive value likely increases the probability of pursuing an HCV evaluation and any category with a negative value likely decreases the probability of receiving an evaluation. Specifically, employed individuals were the most likely to pursue an HCV evaluation. The second and third most influential factors associated with pursuit of an HCV evaluation were the absence of support systems and a history of foster care.

Fig. 2.

Directionality and strength of social and substance use variable influence on linkage-to-care. The sixth principal dimension provides a continuous variable expressed as a linear combination of the categories illustrated. Categories with positive direction increase the likelihood of pursuit of hepatitis C virus (HCV) care, whereas those with negative direction decease the likelihood of pursuing an evaluation. Abbreviations: high school general educational development diploma equivalency, GED; vocational training, VOC

Factors negatively associated with pursuit of an HCV evaluation are also illustrated (Fig. 2). The use of crack/cocaine as the initial illicit substance used was the largest negative predictor of pursuit of an HCV evaluation. The second and third most influential negative predictors were criminal activity without incarceration and the absence of a family history of chemical dependency, respectively.

4. Discussion

Research into SDOH, especially among disadvantaged and underserved populations, has recently been emphasized (Daniel et al., 2018). Healthcare professionals are encouraged to collect SDOH data during routine clinical care interactions; data collection can be facilitated through the EMR, which can subsequently be used for research (Daniel et al., 2018). We sought to understand the influence of social determinants, substance use characteristics, and their interactions on whether a PWSUD would pursue an HCV evaluation. Those most likely to pursue an HCV evaluation were employed with a history of foster care without available social support systems. Conversely, those least likely to link to HCV care reported that their initial illicit substance used was crack or cocaine, that they had been involved in criminal activity without incarceration, and that they had no family history of chemical dependency. Changing HCV demographics, largely as a result of the opioid epidemic, underscores the importance of understanding how social and substance use factors positively or negatively influence HCV linkage to care.

Employment was the strongest factor associated with pursuit of an HCV evaluation. Individuals who are employed are more likely to have health insurance. Given the high cost of HCV medications, being employed facilitates engagement and retention in and successful treatment outcomes not only for substance use disorders (SUD) but also for HCV (Platt, 1995). Employment also confers a variety of secondary benefits including facilitating health care access, greater psychosocial integration, and an increase in educational pursuits (Laudet, 2011; Lusk, 2018). We also found that the absence of support systems and a history of foster care were the second and third most influential variables associated with pursuit of an HCV evaluation, respectively. Individuals with a history of foster care frequently have to contend with health inequalities in the absence of or with minimal social support (notably, without a biological family) (Collins, Jimenez, & Thomas, 2018). While in foster care, youth are accustomed to following directives. Upon exiting foster care, they are expected to be self-sufficient and to navigate the health-care system without substantial social support and frequently with inconsistent living arrangements (Collins et al., 2018; Kruszka, Lindell, Killion, & Criss, 2012). Consequently, individuals with a history of foster care are expected to manage their own health issues. Based upon our observations, once they are part of the health care system, they appear to follow the directions of their providers and other health-care professionals.

In our sample of PWSUDs on methadone, negative predictors of linkage-to-care for HCV were crack or cocaine use as the first illicit substance, participation in criminal activity without incarceration, and no family history of chemical dependency. Cocaine use has been associated with decreased access to healthcare as well as lower rates of linkage to primary, HIV and HCV care (Alavi et al., 2014; Artenie et al, 2015; Bhatia, Hartman, Kallen, Graham, & Giordano, 2011; Gardner et al, 2005; Kenya et al, 2015; Rosen et al, 2013). Among PWSUDs, HCV prevalence is higher among those in the criminal justice system (Marotta, Gilbert, Terlikbayeva, Wu, & El-Bassel, 2018), although only a minority pursues HCV care even when offered monetary incentives (Beckwith et al, 2015; Zaller et al, 2016). These individuals are also more likely to experience poor health outcomes (Dumont, Brockmann, Dickman, Alexander, & Rich, 2012; Morse et al, 2017). Consequently, consistent with our findings, cocaine use and association with the criminal justice system have been shown to be negative predictors of HCV linkage-to-care.

We also observed that no family history of chemical dependency had a negative influence on HCV linkage-to-care. Substantial evidence exists for strong familial aggregation for SUD (Chassin et al, 2016; Merikangas et al, 1998; Sharma, Bruner, Barnett, & Fishman, 2016). People with a family history of chemical dependency characteristically initiate substance use earlier in life, may be less likely to feel stigmatized by receiving treatment in the MTP, and may be more likely to link to HCV care. Of note, family history of chemical dependency had a positive association in our data with HCV linkage-to-care. In contrast, people without a family history of chemical dependency may be characterized by iatrogenic addiction, methadone prescribed as a replacement for opiates initially prescribed for pain. These individuals tend to be older and may be more likely to feel stigmatized receiving treatment in the MTP with its rigorous attendance requirements (Kolodny et al, 2015). These individuals may also be less likely to link to HCV care due to the stigma associated with addiction and its treatment. Of note, the mean age of our patient cohort was 45 years, 10 years older than the mean age of the entire MTP population. Thus, those with iatrogenic opioid addiction may be over represented in our cohort.

Our study fills an important knowledge gap: the role of SDOH, substance use characteristics, and their interactions on linkage-to-care for HCV among PWSUD. We utilized data extracted from the patient’s medical record obtained during routine clinical care, which may provide insights that would have been impossible to obtain from clinical trials. Of note, however, the primary purpose of medical record data is clinical care, not research. As a consequence, data completeness may be an issue; indeed, the statistical analysis and resulting conclusions were limited due to the amount of “no information” present in many categories. However carefully defined, the categories and outcome variables chosen by the investigator may not correspond to the data available in the medical record. Rare categories and rare outcomes have a potential to create imbalanced datasets and groups (e.g., evaluated versus unevaluated patients) and require large sample sizes to permit adequacy of statistical analyses. Thus replication of these results in other, similar populations is important.

Additional advantages of conducting the study in MTP are medical record retention and the completeness of the medical record. The MTP has had an excellent record of maintaining paper medical charts; none have been lost in the past 14 years (personal communication, clinic director). While medical record data confers certain limitations, it is important to note that, compared to data available through a hospital or primary care clinic, SDOH data obtained from an MTP is more complete. For example, the intake assessment at MTPs consists of a comprehensive psychosocial evaluation and inquires about areas such as family and social relationships, legal status, leisure time activities, housing, education level, employment, and mental health diagnosis.

Additional limitations relate to differences in the manner in which data may be captured in the EMR by different providers; inconsistencies in documentation may affect biases in data abstraction. As the data were retrospectively extracted, they may be subject to a variety of patient-reported biases, such as recall and social desirability biases, especially on patient self-reported items that require recollection of information form one’s past (e.g., history of substance use or sexual assault, age of first use, substance of first use, or forensic history). These limitations, however, are inherent to many retrospective investigations that largely rely on self-reported responses. Additionally, confirmation of viral eradication and data collection from more than one MTP would have been highly valuable, but were beyond the study’s scope. The demographics of the study site, however, are representative of other regional MTPs, suggesting that the results may be generalizable to area programs.

4.1. Future directions

Future work should seek to prospectively evaluate our results. The ability to utilize sophistocated statistical techniques to cluster patients according to the likelihood (or lack thereof) to pursue an HCV evaluation could be utilized for resource allocation. These clusters may be utilized by policymakers to enhance the likelihood of entry into HCV care. For example, interventions to increase linkage-to-care could be targeted to clusters characterized by variables highly negatively correlated with HCV treatment pursuit. Similarly, clusters with high expression levels of factors positively associated with treatment pursuit might be prioritized in resource-limited settings.

4.2. Conclusions

Improving treatment uptake in PWSUD is vital to reducing the global burden of HCV (Midgard, Bramness, Skurtveit, Haukeland, & Dalgard, 2016). HCV elimination among PWSUD through participation in HCV care has been emphasized by the recently released Action Plan for Viral Hepatitis (US Department of Health and Human Services, 2017). We found that being employed, the absence of social support, and a history of foster care were the three most influential factors positively associated with pursuit of HCV care. The three factors that were most negatively associated with entry into HCV care were crack/cocaine as the initial illicit substance used, a forensic history without incarceration, and a family history of chemical dependency. An enhanced understanding of these factors in PWSUD is crucial to HCV elimination and can directly contribute to intervention development to promote linkage to HCV care.

Supplementary Material

1

Supplementary Fig 1: Plot of within-cluster sum of squares versus number of clusters for the K-means algorithm applied on the first two principal dimensions of social determinant and substance use variables.

Highlights:

• Social determinants of health (SDOH), such as poverty, education level, and social support, affect health outcomes.

• Data are limited on the influence of SDOH, substance use characteristics, and their interactions on pursuit of HCV care.

• Factors positively (in rank order) associated with pursuit of HCV care were employment, absence of social support systems, and a history of foster care.

• Factors negatively (in rank order) associated with pursuit of HCV care were crack/cocaine as the initial illicit substance, criminal activity without incarceration, and absence of a family history of chemical dependency.

• Identification of subsets of substance users with positive or negative predictors for participation in HCV care could permit more effective resource allocation and enhance treatment uptake.

Abbreviations:

SDOH

Social determinants of health

PWSUD

persons with substance use disorders

HCV

hepatitis C virus

MTP

methadone treatment program

EMR

electronic medical record

WNY

Western New York

OUD

opioid use disorder

MCA

multiple correspondence analysis

LASSO

least absolute shrinkage and selection operator

SUD

substance use disorders