Perceived Behavioral Control and Barriers to Cleaning Skin Before Injecting Drugs

Abstract

Objectives:

Skin and soft tissue infections (SSTI) among people who inject drugs (PWID) are common and represent a significant public health burden. In the current study, we examined the relationship between perceived behavioral control and perceived barriers to cleaning skin prior to injecting drugs.

Methods:

Participants (n = 248; 37.9 (± 10.7) years of age, 58.5% male, 59.3% white, 16.1% Hispanic) were patients seeking medical care at a large urban hospital in the northeastern United States. We used ordinary least squares regression to estimate the associations between perceived barriers to skin cleaning with background characteristics and perceived behavioral control.

Results:

Controlling for background and other study variables, greater number of past year skin abscesses was associated with greater level of perceived barriers to skin cleaning (p < .001), while higher level of education and higher perceived behavioral control were associated with lower levels of perceived barriers to skin cleaning (p < .001).

Conclusions:

Interventions aiming to reduce the likelihood for SSTI among people who inject drugs may benefit from strengthening individual’s behavioral control and providing skin cleaning skills training.

1.0. INTRODUCTION

People who inject drugs (PWID) face significant risk for adverse consequences, including overdose, HIV, hepatitis C, and bacterial infections (e.g., skin infections, abscesses, sepsis, and endocarditis). Skin and soft tissue infections (SSTI) are particularly common among PWID; nearly one in three PWID in a San Francisco community-based sample report a current skin infection (Binswanger et al. 2000) and rates for recent SSTI among PWID internationally range from 10% (current SSTI) to 20% (SSTI in the past six months) (Lloyd-Smith et al. 2008, Pollini et al. 2010). In the U.S., hospitalization rates for opiate-related SSTI doubled from 1993 to 2010 (Ciccarone et al. 2016). SSTI are among the most common reasons for emergency department visits, and PWID with SSTI represent a substantial burden for hospital systems, particularly in urban hospitals in the West and Northeast (Takahashi et al. 2010, Tookes et al. 2015).

Cleaning one’s skin with alcohol prior to injection minimizes risk for SSTI, yet many PWID fail to regularly do so (Murphy et al. 2001, Smith et al. 2015). Common barriers to skin cleaning include drug craving and withdrawal, already being high at the time of injection, failure to carry alcohol or alcohol wipes, social network rituals, and time constraints (Bonar and Rosenberg 2014, Phillips 2016). PWID who perceive greater barriers to skin cleaning are less likely to implement protective skin cleaning behaviors (Bonar and Rosenberg 2014, Phillips 2016). Although self-efficacy is associated with lower injection risk behavior (for review see Wagner et al. 2010), no studies to date have examined how cognitive factors contribute to one’s perceptions about barriers to skin cleaning, which may inform ways to increase PWID’s use of skin cleaning practices.

Grounded in the theory of planned behavior (TPB; Jozaghi and Carleton 2015), perceived behavioral control encompasses one’s perceived ability or difficulty to effectively perform a particular behavior to prevent an undesired outcome. Behavioral control has emerged as a primary predictor of behavioral intention and future behavior (e.g., Povey et al. 2000, Guo et al. 2007, Hunt and Gross 2009, Chabot et al. 2010), including intention to share needles among PWID (Jozaghi et al. 2016). Accordingly, PWID who feel they have control over a situation should be better able to implement healthy behaviors (i.e., cleaning skin prior to injection) and avoid harmful outcomes (e.g., SSTI).

Several demographic variables have been associated with injection risk behaviors and SSTI. Although Phillips and colleagues (2017) did not find past year SSTI associated with sex, race/ethnicity, educational attainment or homelessness, other studies had found significant associations between increased risk for SSTI and both being male and homeless (Pollini et al 2010). Years of injection drug use (IDU) is correlated with lower rates of SSTI and other bacterial infections (Binswanger et al. 2000, Phillips et al. 2008), indicating that PWID with significant experience may engage in protective behaviors to minimize perceived barriers to cleaning skin and reduce overall risk for SSTI.

In the current study, we hypothesized that greater perceived behavioral control would be associated with lower levels of perceived barriers to cleaning skin prior to injection among PWID. We controlled for past year SSTI, which was expected to be associated with greater perception of barriers, and other background variables (age, sex, race, educational status, homelessness, years of IDU) associated with SSTI.

2.0. Materials and Methods

2.1. Recruitment

Between January 2014 and August 2018, we recruited patients from inpatient units at an academic safety-net hospital as part of an ongoing randomized trial targeting skin-cleaning education and other risk reduction strategies for PWID. The Boston Medical Center Institutional Review Board approved all materials and recruitment procedures used in this study. Study research staff reviewed electronic medical records daily to identify patients showing evidence of illicit drug use. After receiving permission from the clinical team, research assistants screened interested patients and invited those eligible to take part in the study.

Eligibility criteria included being 18 years or older and self-reporting IDU at least three times in the week prior to hospitalization. Participants reporting current psychosis, suicidality, or who could not speak English or otherwise provide consent, could not provide the names for at least two contact persons, or planned to move out of Boston in the next year were excluded. Of the total 938 patients screened, 566 (60.3%) were deemed ineligible: no drug injection (n = 288), injected less than required (n = 114), no contacts (n = 54), did not speak English (n=17), and other reasons (n = 93). Among those eligible (n = 372), 71 (19.1%) refused to participate and 49 were discharged prior to consent. The remaining 252 participants consented to participate and completed the baseline assessment. Of these, 248 completed all study measures. Data for the current study came from the baseline assessment, which consisted of a structured 60–90 minute in-room interview. Participants were compensated with the choice of either a $20 gift card or pre-paid cell phone for use until study completion (Stewart et al. 2018).

2.2. Measures

In addition to age, sex, race, ethnicity, and years of education the following variables were assessed.

2.2.1. Homelessness.

Respondents were asked where they had slept/spent their nights in the past three months. Respondents spending any nights on the street or in a shelter were coded as homeless.

2.2.2. Injection drug use.

Years of IDU was calculated by subtracting respondents’ current age by their reported age at which they first injected drugs. Respondents were also asked how many times they had injected heroin, cocaine or other drugs in the past 90 days. All injections were summed to form a variable assessing total injections in the past 90 days.

2.2.3. Number of SSTI in the past year.

Respondents were asked how many times they have had an abscess or skin infection in the past year. The following definition was provided: “Skin infections include abscesses (red, hardish, infected lumps that contain pockets of pus), ulcers (open infected sores that look like a crater), and cellulitis (a more widespread skin infection) that occur at the infection site.”

2.2.4. Perceived behavioral control.

Respondents were asked how strongly they agreed or disagreed (six-point Likert) with five statements (e.g., “I have little control over the things that happen to me”) (Whitaker et al. 2000). A summed composite was used, with higher values denoting higher levels of behavioral control. Although this measure has not been used to assess perceived control among PWID, its internal consistency reliability was high (Chronbach’s α = .93) in this sample.

2.2.5. Barriers to skin cleaning.

Respondents were asked how much they agreed or disagreed with each of twenty “reasons that make it hard for you to clean your skin” before injecting a drug of choice (e.g., “Cleaning my skin before injecting interrupts the ritual of using.”). Items were developed for a recent study (Phillips 2016) and are based on clinical work with PWID and two qualitative studies (Gleghorn and Corby 1996, Treloar and Cao 2005). Response options ranged from strongly disagree to strongly agree (five-point Likert), and responses were summed to form a composite (Chronbach’s α = .89 in the present sample).

2.3. Analytical Methods

We present descriptive statistics to summarize the characteristics of the sample. Chronbach’s alpha was used to evaluate the internal consistency reliability of constructed indexes. We used ordinary least squares regression to estimate unadjusted and adjusted associations between perceived barriers to skin cleaning with background characteristics and perceived control. Because residuals exhibited heteroscedasticity, we used the Huber-White variance estimator to estimate 95% confidence intervals and test the statistical significance of associations. Because variables are on different metrics we report y-standardized coefficients for categorical covariates and fully standardized coefficients for continuous covariate to facilitate interpretation. Unless inconsistent we discuss only the adjusted coefficients estimated by multiple linear regression in the text. All analyses were conducted using Stata 15.1 (StataCorp 2017).

3.0. Results

Participants averaged 37.9 (± 10.7) years of age, 58.5% were male, 59.3% were white, 20.6% were Black, 20.2% were of other racial origins, and 16.1% were Hispanic (Table 1). Mean educational attainment was 11.6 (± 2.43) and 62.1% had experienced homelessness in the past 90 days. Mean years of injection drug use was 12.3 (± 11.1, Median = 9) and participants reported an average of 359.0 (± 416.7, Median = 255) injections in the past 90 days. In all, 98.0%, 60.1%, and 24.2% of participants reported injecting heroin, cocaine, and other drugs in the past 90 days, respectively. Lifetime number of reported skin abscesses ranged from 0 – 300; mean = 5.37 (± 20.8, Median = 2.) and 79.4% reported at least one prior skin infection.

Table 1.

Background Characteristics (n = 248).

	Mean (± SD) or n (%)
Years Age (Mean, ± SD)	37.9 (± 10.7)
n (%) Male	145 (58.5%)
Race
n (%) White	147 (59.3%)
n (%) Black	51 (20.6%)
n (%) Other	50 (20.2%)
n (%) Hispanic	40 (16.1%)
Years Education (Mean, ± SD)	11.6 (± 2.43)
n (%) Any Homelessness	154 (62.1%)
n (%) Past 90-Day Heroin IDU	243 (98.0%)
n (%) Past 90-Day Cocaine IDU	149 (60.1%)
n (%) Injected Other Drugsa	60 (24.2%)
Total # of Injections (Mean, ± SD)	359.0 (± 416.7)
Years Inj. Drug Use (Mean, ± SD)	12.3 (± 11.1)
n (%) Any Skin Infections (Lifetime SSTI) (Yes)	197 (79.4%)
Past Year Skin Infections (Mean, ± SD)	5.37 (± 20.8)
Barriers to Skin Cleaning (α = .888)b	2.90 (± 0.78)
Perceived Control (α = .930)c	3.78 (± 1.04)

^aIncludes 29 persons who said they had injected methamphetamines, 17 who said they had injected opioids other than heroin, and 14 who said they had injected both methamphetamines and opioids other than heroin.

^bResponse categories ranged from 1, strongly disagree, to 5, strongly agree. The index score was generated as the mean score of items comprising the index. Median = 3. Cronbach’s α = .888.

^cResponse categories ranged from 1, strongly agree, to 6, strongly disagree. The index scores was generated as the mean score of 5-items comprising the index. Higher scores indicate higher perceived control. Cronbach’s α = .930.

Table 2 gives unadjusted and adjusted associations of barriers to practicing skin cleaning with background characteristics and perceived control. The unadjusted associations of educational attainment (βy.x = −0.17, p = .010), homelessness (βy.x = 0.39, p = .005) and total injections in the past 90 days (βy.x = 0.17, p = .009) were statistically significant, though these associations were not significant at the .05 level after control for other covariates in the multiple linear regression model. Perceived barriers to practicing skin cleaning was positively and significantly (adjusted βy.x = 0.14, p < .001) associated with number of skin infections and inversely and significantly associated with perceived behavioral control (adjusted βy.x = −0.25, p < .001).

Table 2.

Unadjusted and Adjusted Association of Barriers to Practicing Skin Cleaning with Background Characteristics and Perceived Control (n = 248).

	UNADJUSTED		ADJUSTED
	βy.xa (95% CI)	p =	βy.xa (95% CI)	p =
Years Age	−0.12 (−0.30; 0.00)	.054	−0.08 (−0.27; 0.11)	.395
Sex (Male)	−0.01 (−0.28; 0.26)	.964	0.07 (−.0.23; 0.38)	.637
Non-Hispanic White (Yes)	0.15 (−0.12; 0.42)	.277	0.14 (−0.15; 0.43)	.337
Years Education	−0.17 (−0.29; −0.04)	.010	−0.10 (−0.23; 0.02)	.097
Homelessness (Yes)	0.39 (0.12; 0.67)	.005	0.26 (−0.02; 0.52)	.070
Years IDU	−0.03 (−0.15; 0.09)	.633	−0.03 (−0.17; 0.12)	.726
Log(Total Injections)	0.17 (0.04; 0.30)	.009	0.13 (−0.00; 0.26)	.053
# Skin Abscesses	0.14 (0.08; 0.21)	<.001	0.14 (0.07; 0.21)	< .001
Perceived Control	−0.25 (−0.30; 0.04)	<.001	−0.28 (−0.39; −0.14)	<.001

^aAssociations were estimated by ordinary least squares regression. Coefficients are fully-standardized for continuous covariates and are y-standardized for categorical covariates. Confidence interval estimates and tests of significance were based on the robust Huber-White variance estimator.

4.0. Discussion

SSTI are common among PWID and these results shed light on factors associated with PWID’s perceptions about barriers to skin cleaning, which is thought to be a key risk factor for such infections. Consistent with hypotheses, accounting for other variables, PWID reporting lower behavioral control reported greater barriers to skin cleaning. Further, number of past year SSTI was associated with greater perceived skin cleaning barriers. In addition, although educational status was inversely associated and homelessness positively associated with perceived barriers, the association between homelessness and perceived barriers was attenuated in the full model.

Consistent with TPB (Jozaghi and Carleton 2015), in this sample, PWID with greater levels of perceived behavioral control endorsed fewer barriers to skin cleaning, which may enable them to more effectively implement skin cleaning behaviors to avoid harmful outcomes. Deficits in behavioral control are particularly risky for individuals who perceive many barriers to cleaning skin within environments laden with drug-using stimuli. Cognitive-behavioral interventions targeted at helping PWID identify triggers for injecting unsafely and developing skills for overcoming perceived barriers to cleaning skin are needed.

These findings linking recent SSTI with beliefs about barriers that prevent PWID from cleaning skin prior to injection may partially explain why PWID experience recurring SSTI. Although years of IDU has been correlated with lower rates of SSTI and other bacterial infections (Binswanger et al. 2000), it was not associated with perceived barriers to skin cleaning in the present study. Therefore, rather than assessing years of IDU when assessing PWID’s risk for SSTI it may be more important to account for PWID’s history of SSTI, which may be a proxy for unsafe injection behaviors. Further, it may be useful to target PWID diagnosed with SSTI when they present to the emergency department or inpatient hospitalization to offer skin cleaning and other risk reduction training. Such training is likely to be most effective when offered in conjunction with counseling (e.g., motivational interviewing) about how to successfully overcome perceived barriers.

Although homelessness was significantly correlated with perceived barriers to skin cleaning, this relationship was fully attenuated after adjusting for perceived control and background factors. Surprisingly, despite environmental and financial challenges and the heightened risk for contracting HIV among public injectors (German et al. 2007, Harris et al. 2018), homeless PWID in this sample did not perceive greater barriers to practicing pre-injection skin cleaning relative to non-homeless PWID once covariates were included in the model. More research is needed to examine how perceived control may interact with homeless PWID’s skin cleaning practices.

PWID with higher levels of education may perceive lower levels of barriers to skin cleaning because they have had greater exposure to harm reduction strategies or may better access to such strategies. While risk reduction education should account for PWID’s current level of awareness and knowledge, future research is needed to better understand which barriers can be remedied through broader public health initiatives (e.g., access to alcohol or alcohol wipes; concerns about arrest).

4.1. Limitations

The current analyses did not assess some TPB constructs, including attitudes toward implementing skin cleaning and subjective norms (i.e., one’s perceptions about others’ beliefs about cleaning or not cleaning skin prior to injection). Also, this study assessed global behavioral control rather than behavioral control specific to safe injection behaviors (e.g., When craving sometimes you don’t really care about risk of sharing; Jozaghi et al. 2016). Studies examining the impact of behavioral control in the context of skin cleaning and risky injection use may be warranted. Moreover, given that failed injection attempts and sites of injection (e.g., intramuscular, femoral vein) (Phillips and Stein 2010, Ciccarone and Harris 2015) may exacerbate risk for bacterial skin infection, studies that assess where PWID inject on their bodies and account for failed injection attempts specifically may shed light on PWID’s injection-related challenges and risk. Moreover, given the escalated risks associated with public versus private injection behaviors (e.g., Harris et al. 2018), gaining insight into injection locations is warranted. Finally, regardless of skin cleaning, biological factors such as skin microbiome may play a role in the incidence of skin infections.

4.2. Conclusions

SSTI are a primary contributor to morbidity and health care costs in the opioid addiction epidemic. Interventions aiming to reduce the likelihood of SSTI among people who inject drugs may benefit from strengthening individual’s behavioral control and providing skin cleaning skills training. Such interventions could be delivered during hospitalization or through emergency departments, at needle exchanges, or perhaps in the future at safe injection facilities, although entry into drug treatment and reduction in injection frequency remain key elements in harm reduction for this population (Stein and Anderson 2003).