Overdose response among trained and untrained women with a history of illicit drug use: a mixed-methods examination

Abstract

Little is known about differences in bystander behavior among people who use drugs, trained and untrained in opioid overdose prevention. We examined three types of recommended overdose response - a 911 call, rescue breathing/CPR, and naloxone administration—among Philadelphia-based, predominantly street-involved women with a history of illicit drug use. The study utilized a convergent mixed methods approach integrating data from 186 quantitative survey responses and 38 semi-structured qualitative interviews. Quantitative findings revealed that compared to untrained women, trained women were more likely to administer naloxone (32.9% vs. 5.2%) and use two recommended responses (20.0% vs. 9.5%). No significant differences were found between the two groups in calling 911 or using rescue breathing/CPR. Qualitative findings indicated that barriers to enacting recommended overdose response were either structural or situational and included the avoidance of police, inability to carry naloxone or phone due to unstable housing, and perceived lack of safety on the streets and when interacting with strangers. Our study demonstrated that overdose training improved the frequency of naloxone administration among this sample of predominantly street-involved women. Future efforts need to focus on avoiding intrusive policing, scaling-up naloxone refill sites, and providing secondary naloxone distribution via drug user networks.

Introduction

The steep increase in opioid-involved overdose deaths in the United States in the last decade—from 21,088 deaths in 2010 to 46,802 deaths in 2018 (National Institute on Drug Abuse, 2020)—has been paralleled by a wide-spread expansion of community-based overdose prevention training programs. From 2010 to 2014, the number of overdose training sites offering naloxone, an opioid antidote also known by the trade name Narcan, increased from 188 in 15 U.S. states and District of Columbia (DC) to 644 in 30 states and DC (Wheeler et al., 2012, 2015). Though the administration of naloxone is a key element of opioid overdose reversal, other important actions are also involved. According to the Substance Abuse and Mental Health Services Administration (SAMHSA), the five essential steps of responding to an opioid overdose include checking for overdose signs, calling 911, administering naloxone, supporting the person’s breathing, and monitoring the victim (SAMHSA, 2018).

People who use drugs witness multiple overdoses in their life, with the median ranging from two (Bohnert et al., 2012) to six (Sherman et al., 2008). Research shows that overdose prevention trainings that target people who use drugs increase their knowledge of appropriate overdose responses (Wagner et al., 2010), improve self-efficacy to respond to overdose situations (Ashrafioun et al., 2016; Bennett & Holloway, 2012; Lankenau et al., 2013; Maldjian et al., 2016; Pietrusza et al., 2018; Wagner et al., 2014), and, most importantly, contribute to the use of naloxone in overdose situations (Bennett et al., 2011; Wagner et al., 2010). Still, a number of individual and structural factors, including intoxication, panic, as well as fear of police involvement and arrest (Lankenau et al., 2013; Sherman et al., 2008), can hinder an appropriate overdose response among witnesses who actively use drugs.

While prior research has been largely focused on the improvement of knowledge and behaviors among people who use drugs who have received overdose prevention training, scarce evidence exists on how trained and untrained individuals may respond differently to a drug overdose. One prospective evaluation of a UK-based overdose prevention program (Bennett & Holloway, 2012) compared 28 trained and 38 untrained drug users and found that trainees were more likely to call an ambulance and place the victim in the recovery position, but less likely to use CPR (cardiopulmonary resuscitation) than untrained participants. Two other studies, however, did not find substantial differences in the utilization of recommended overdose response between the two groups. A qualitative examination of 22 trained and 9 untrained people who inject drugs reported that naloxone was used mainly by trained participants at 58% of last witnessed overdoses, however in other cases, other life-saving measures (such as a 911 call or CPR) were also successfully used by both groups (Sherman et al., 2008). Likewise, a retrospective cohort study among 373 substance-using people who administered naloxone during overdose rescue found no significant differences in the behaviors of overdose rescuers pre- and post-training (Doe-Simkins et al., 2014).

The current study focuses on differences in overdose witnessing behaviors between trained and untrained women with a history of illicit drug use. While women’s overdose response has been largely overlooked by prior research, limited evidence suggests that women may carry naloxone more often than men (Tobin et al., 2018) and are more likely than men to seek help in overdose situations, e.g., calling 911 (Ambrose et al., 2016; Lim et al., 2019). Women may also witness particular types of overdose. Since women who inject drugs are more likely than men to have sexual partners who also inject drugs (Evans et al., 2003; Fals-Stewart et al., 2003; Powis et al., 1996), women may carry a particular burden of witnessing and responding to an overdose in their close relationships. Also, women who use drugs may witness a substantial number of overdoses on the streets given a co-occurrence of women’s illicit drug use with different forms of street involvement, including unstable housing (Elifson et al., 2007), street-based drug sales (Miller & Neaigus, 2002), or street-based sex work (Chettiar et al., 2010; Miller et al., 2011). Yet, it is unclear whether a potentially high exposure to street-based overdoses can elicit a high level of response from drug-using women who may be preoccupied with the hardships of immediate survival on the streets (Epele, 2002).

We pursued two aims in this study: (1) to examine the associations between women’s overdose prevention training and the use of recommended techniques at the last witnessed overdose, including a 911 call, the administration of naloxone, and rescue breathing/CPR; and, (2) to identify contextual factors surrounding women’s overdose response.

Methods

Mixed-methods design

We employed a convergent mixed methods design (Creswell & Plano Clark, 2017), collecting quantitative and qualitative data in parallel and integrating them at the analysis stage to gain a complex understanding of factors, including overdose training, involved in women’s bystander behavior. A survey gathered quantitative data to examine associations between training exposure and overdose response while qualitative interviews contextualized quantitative findings and identified barriers to recommended overdose responses.

Study location

The study utilized venue-based recruitment from sites served by Prevention Point Philadelphia (PPP), a Philadelphia-based harm reduction agency that serves vulnerable populations, including people who use drugs. The organization is located in Kensington, an impoverished neighborhood in North Philadelphia marked by a large concentration of open-air drug sales (Fairbanks, 2011) and a high density of opioid-related overdose deaths (Philadelphia Department of Public Health, 2018). PPP has offered overdose prevention trainings since 2006. A training session typically consists of learning the basic chemistry of opiates, overdose risk factors and symptoms, and rescue techniques. Participants are also trained in how not to respond with folk remedies, such as injecting milk or salt. Upon training completion, participants are provided with overdose rescue kits, which include two 4 mg bottles of intranasal naloxone, rubber gloves, and a face shield for performing rescue breathing.

Sampling

Data were drawn from a larger cross-sectional study examining personal and witnessed overdose among women who accessed PPP for any service. Eligibility criteria included age 18 or older, being non-pregnant, and the ability to read. Between January 2016 and January 2017, 220 women were enrolled, including 199 women recruited at the PPP main office in Kensington and 21 women recruited at five mobile harm reduction points in Philadelphia. Compared to the women recruited from the PPP main office, those recruited at mobile sites were older and more likely to be non-White; however, the groups did not significantly differ on the exposure (overdose training status) or key outcomes, such as calling 911, administering naloxone, or performing recue breathing/CPR at the last witnessed overdose.

A subgroup of women (n = 42) were purposely selected from the quantitative sample (n = 220) for semi-structured qualitative interviews to provide narrative accounts of personally experiencing or witnessing an overdose. The selection criteria included: witnessing at least one overdose in the past 12 months, having at least one opioid injection (heroin or prescription opioid) in the past 30 days, living in Philadelphia area, and being recently street-involved. Qualitative interviews, which followed the quantitative survey, were conducted within several hours or up to several months of the quantitative survey depending upon the availability of women selected for the qualitative examination.

The analytical sample for the quantitative component of this study is 186 women. Since the present analysis focuses on women’s responses to a witnessed overdose, 25 participants who have never witnessed an overdose were excluded from the total sample of 220 participants. Additionally, we excluded eight women with an unknown training status before the most recently witnessed overdose and one woman with missing data on her witnessing behavior at the most recent overdose. The analytical sample for the qualitative part of this study is 38 women. Among the 42 participants who were originally selected for qualitative interviews, 4 could not describe their response at the last witnessed overdose or identify the last witnessed episode and were excluded.

The study was approved by Drexel University IRB. All participants provided verbal informed consent before taking part in the quantitative survey or qualitative interview.

Quantitative data collection

Procedures

Participants were surveyed with a 97-item, interviewer-administered questionnaire. Survey locations included the PPP main office, mobile van, or other mutually agreed location, such as a nearby fast-food restaurant. Data were collected using Qualtrics survey software (Provo, UT, 2018). The survey took between 30 and 40 min. Multiple-choice questions were accompanied by paper cards, which showed available response options; the participants looked at the cards and read back applicable responses. Participants received $10 in cash upon survey completion.

Measures

Sociodemographics and drug use.

Women were asked about age, race, Hispanic origin, cell phone ownership, as well as their past 30-day drug use practices, including heroin and fentanyl use.

Recent street involvement.

Since an inclusive definition of street involvement encompasses both living on the streets as well as spending substantial time on the streets or participating in street economy (DeMatteo et al., 1999; Lankenau et al., 2005; Worthington & MacLaurin, 2009), recent street involvement was determined by one or more of the following activities: past 30-day drug use or drug sales in public places, past 12-month survival sex (exchanging sex for money, drugs, housing, or gifts), or past 12-month unstable housing. Unstable housing was determined by endorsing any of the following items: living in a car, abandoned building/house, shelter, motel, transitional housing, on the streets, temporarily living with friends or family, or having any other unstable housing.

Lifetime personal and witnessed overdose.

The frequency of lifetime personal overdoses was measured with a single question, In your lifetime, how many times have you overdosed on drugs? By ‘drug overdose’ I mean a situation when you’re taking more of a drug than your body can process. It can lead to some serious problems like blackout, inability to breathe or even heart failure. The frequency of lifetime witnessed overdoses was measured with two questions: In your lifetime, how many times did you witness a drug overdose (fatal or non-fatal)? and How many of them were fatal overdoses?

Circumstances of the most recently witnessed overdose.

Women were asked about the timing of the last witnessed overdose, overdose victim, and overdose place. To examine whether participants primarily witnessed opioid overdoses, they were asked, How did you recognize that overdose? This question was followed by multiple choice items describing opioid overdose symptoms, such as being unresponsive, unable to breathe, having bluish skin or fingernails, being unconsciousness or in a coma, and others; an open-ended answer option was offered as well.

Response at the last witnessed overdose (dependent variable).

Participants were asked about their response to the most recently witnessed overdose; multiple-choice options are listed in Table 2. Three types of overdose response were identified as recommended responses: 911 call, rescue breathing/CPR, and naloxone use. While rescue breathing is a component of CPR (American Heart Association, 2020), we asked about the use of the both procedures since participants frequently used the terms ‘rescue breathing’ and ‘CPR’ interchangeably. Participants who did not administer naloxone during the last witnessed overdose were asked, You didn’t administer naloxone during the most recent overdose that you witnessed, why? (answer options included: being trained after witnessing that overdose, other people administered naloxone, not carrying naloxone, and an open-ended response). In addition, composite variables for combinations of three recommended responses were calculated, including: at least one of three recommended responses, only one recommended response, only two recommended responses, and all three recommended responses (Table 3).

Table 2.

Response at the last witnessed overdose, by overdose prevention training.

Type of response	Total (N = 186) % (n)a	Trained (n = 70) % (n)a	Untrained (n = 116) % (n)a	p-Value*
Recommended response
911 call	48.4 (90)	48.6 (34)	48.3 (56)	.97
Naloxone use	15.6 (29)	32.9 (23)	5.2 (6)	<.001
Rescue breathing/CPR	14.5 (27)	11.4 (8)	16.4 (19)	.35
Additional response
Slapped, hit or pinched the person	19.9 (37)	18.6 (13)	20.7 (24)	.73
Put water in the person’s face	18.8 (35)	17.1 (12)	19.8 (23)	.65
Kept the person awake	13.4 (25)	5.7 (4)	18.1 (21)	.02
Put the person in a cold bath or shower	2.2 (4)	1.4 (1)	2.6 (3)	.60
Took the person to the ER/hospital	1.1 (2)	0	1.7 (2)	.27
Injected the person with milk	0.5 (1)	1.4 (1)	0	.20
Injected the person with salt	0.5 (1)	0	0.9 (1)	.44
Injected the person with another drug	0.5 (1)	0	0.9 (1)	.44
Other actionsb	19.4 (36)	18.6 (13)	19.8 (23)	.83
Lack of any responsec	17.7 (33)	12.9 (9)	20.7 (24)	.18

^aResponse percentages can exceed 100% because participants could select multiple answers.

^bIn open-ended answers following the endorsement of ‘Other’ category, 3 participants indicated they used sternal rub, 12 participants sought help from others, and 21 participants were involved in other activities.

^cIn open-ended answers following the endorsement of ‘Did nothing’ category, 21 participants indicated that the lack of action was related to the involvement of other people in overdose response.

^*Bolded p-values (<.05) indicate significant differences between trained and untrained groups.

Table 3.

Combinations of recommended responses at the last witnessed overdose, by overdose prevention training.

Recommended response	Total (N = 186) % (n)	Trained (n = 70) % (n)	Untrained (n = 116) % (n)	p-Value*
At least one recommended response
911 call or naloxone use or rescue breathing/CPR	64.0 (119)	72.9 (51)	58.6 (68)	.05
Only one recommended response
911 call (no rescue breathing/CPR or naloxone)	35.5 (66)	31.4 (22)	37.9 (44)	.37
Naloxone use (no 911 or rescue breathing/CPR)	10.2 (19)	20.0 (14)	4.3 (5)	.001
Rescue breathing/CPR (no 911 or naloxone)	4.3 (8)	1.4 (1)	6.0 (7)	.13
Total for one recommended response	50.0 (93)	52.9 (37)	48.3 (56)	.55
Two recommended responses
911 call and rescue breathing/CPR (no naloxone)	8.6 (16)	7.1 (5)	9.5 (11)	.58
911 call and naloxone use (no rescue breathing/CPR)	3.8 (7)	10.0 (7)	0	.001
Naloxone use and rescue breathing/CPR (no 911 call)	1.1 (2)	2.9 (2)	0	.07
Total for two recommended responses	13.4 (25)	20.0 (14)	9.5 (11)	.04
Three recommended responses
911 call, naloxone use, and rescue breathing/CPR	0.5 (1)	0	0.9 (1)	.44

^*Bolded p-values (<.05) indicate significant differences between trained and untrained groups.

Overdose prevention training.

Exposure to training was assessed with the following question, How many times were you trained in overdose prevention? Timing of overdose training was assessed with two questions: When did you receive your first overdose prevention training? and When did you receive your most recent (or the only) training? A partial subset of trained women (n = 38) received two additional questions, which were included in the survey after the start of data collection, asking about a training venue (PPP or other location) and frequency of carrying naloxone (responses ranged from ‘always’ to ‘never’).

Training status before witnessing the most recent overdose (independent variable).

The participant was considered ‘trained’ if (1) the first, recent, or the only overdose prevention training preceded the most recently witnessed overdose, or (2) if the dates of the most recent training and most recently witnessed overdose coincided, but the participant was trained 2 or more times. The participant was considered ‘untrained’ if she was trained after the last witnessed overdose or never trained. If it was not possible to determine temporality, the training status was marked as ‘unknown’ and the participant was excluded from the analytical sample.

Qualitative data collection

Qualitative interviews lasted up to 90 min and took place in the Prevention Point Philadelphia’s office. Since the larger study aimed to understand the participants’ experiences of personal and witnessed overdose through the realities of their past and present lives, the interview guide included semi-structured questions about participants’ life trajectory, health, social circle, drug use and treatment experiences, personal and witnessed overdose, and overdose prevention training. The major question related to witnessed overdose was, Please describe the most recent case of witnessing a drug overdose. The question was followed with probes asking more details about actions undertaken, place, victim, and other bystanders’ response. The interviews were recorded with a digital voice-recorder. Women were compensated with $25 in cash for their time.

Data analysis

Quantitative analysis

Quantitative analyses were conducted in IBM SPSS Statistics for Windows, version 25 (IBM Corp, 2017). Sociodemographics, drug use, overdose characteristics, and overdose response were stratified by training status. To examine the statistical significance of differences between trained and untrained women, we used: (1) t-test to compare the means of age; (2) Mann-Whitney U test to compare the medians of lifetime personal and witnessed overdoses; and (3) logistic regression and Chi-square test to compare proportions for all categorical variables.

Qualitative data analysis

Qualitative interviews were transcribed, converted into Microsoft Word documents, and exported to MAXQDA 2018 (VERBI Software, 2018). A pre-existing coding scheme was developed in accordance to the interview guide and pretested with the involvement of the second coder. The qualitative dataset used for this study consisted of interview excerpts with a priori code ‘last witnessed overdose.’

The analysis of qualitative data was guided by the semantic/explicit thematic approach (Braun & Clarke, 2006) and consisted of several steps. First, interviews were coded in a deductive way to capture any of three recommended responses during the last witnessed overdose: 911 call, rescue breathing/CPR, or naloxone administration. We coded the use of three recommended responses within each interviewee and counted the number of interviewees who utilized each recommended response. Second, we examined patterns of those responses, including their typical sequence and combination, and perceived barriers to use; we also inductively coded for emerging themes, for example, ‘sources of naloxone.’ Third, based on a priori code ‘overdose training,’ we assigned a training status to the interviewees (trained or untrained before the last witnessed overdose) and compared the two groups on the use of recommended overdose responses.

Results

Sociodemographics, drug use, and lifetime overdose

Sociodemographic, drug use, and overdose characteristics of survey participants, including the stratification of results by the training status, are presented in Table 1. The mean age was 38.6 years and 64% were White. Almost 90% were recently street-involved and only 58% owned a cell phone.

Table 1.

Sociodemographic, drug use, and overdose characteristics of survey participants, by overdose prevention training.

Characteristic	Total (N = 186) % (n)	Trained (n = 70) % (n)	Untrained (n = 116) % (n)	p-Value*
Age, mean (SD)a	38.6 (9.8)	37.8 (9.2)	39.2 (10.2)	.36
Race (any ethnicity)
Caucasian/White	64.0 (119)	67.1 (47)	62.1(72)	ref
African-American/Black	19.4 (36)	17.1 (12)	20.7 (24)	.51
Hispanic	8.6 (16)	8.6 (6)	8.6 (10)	.88
Native American or Pacific Islander	2.2 (4)	2.9 (2)	1.7 (2)	.68
Two or more races	5.9 (11)	4.3 (3)	6.9 (8)	.43
Hispanic origin (any race)b	12.5 (23)	12.9 (9)	12.3 (14)	.91
Recent street involvementa	89.7 (166)	92.8 (64)	87.9 (102)	.30
Unstable housing, past 12 months	71.5 (133)	70.0 (49)	72.4 (84)	.72
Survival sex, past 12 months	58.6 (109)	58.6 (41)	58.6 (68)	.995
Drug use in public places, past 30 days	72.6 (135)	80.0 (56)	68.1 (79)	.08
Drug sales in public places, past 30 daysc	28.0 (51)	32.4 (22)	25.4 (29)	.32
Drug use in the past 30 days
Injection	69.9 (130)	80.0 (56)	63.8 (74)	.020
Any opioid use	79.6 (148)	90.0 (63)	73.3 (85)	.006
Heroin	75.3 (140)	87.1 (61)	68.1 (79)	.004
Cocaine or crack	72.0 (134)	71.4 (50)	72.4 (84)	.89
Mixing any opioid with any benzodiazepine	81.2 (151)	90.0 (63)	75.9 (88)	.017
Ownership of a cell phoned	58.1 (100)	58.2 (39)	58.1 (61)	.99
Lifetime personal overdose
Ever had a personal overdose	70.4 (131)	71.4 (50)	69.8 (81)	.82
Number of personal overdoses, median (IQR)	2.0 (0–4)	2.5 (0–5)	2.0 (0–4)	.62
Lifetime witnessed overdose
Ever witnessed a fatal overdosee	55.5 (91)	62.3 (38)	51.5 (53)	.18
Number of witnessed overdoses, median (IQR)	5.0 (3–10)	9.5 (4.8–20)	5.0 (2–10)	<.000
Most recently witnessed overdose
Timing
In the past year or earlier	77.4 (144)	95.7 (67)	66.4 (77)	.000
Location
Outdoor placesf	61.8 (115)	77.1 (54)	52.6 (61)	.017
Private residenceg	23.1 (43)	15.7 (11)	27.6 (32)	ref
Abandoned buildings or drug use ‘houses’	8.1 (15)	2.9 (2)	11.2 (13)	.34
Car	3.8 (7)	1.4 (1)	5.2 (6)	.52
Other private placesh	3.2 (6)	2.9 (2)	3.4 (4)	.69
Victim
Stranger	58.1 (108)	78.6 (55)	45.7 (53)	.018
Friend	32.8 (61)	17.1 (12)	42.2 (49)	.85
Intimate partner or family member	9.1 (17)	4.3 (3)	12.1 (14)	ref

^a-eSuperscripted letters indicate variation in sample size due to missing responses, ‘Do not know’ responses, or modification of survey instrument: ^an = 185; ^bn = 184; ^cn = 182; ^dn = 172; ^en = 164.

^fStreets, alleyways, parks, parking lots, playgrounds, abandoned areas or lots, railroad tracks, and woods.

^gThe participant’s, victim’s or someone else’s place.

^hRecovery house (2 responses), a store (2 responses), methadone clinic, and nonprofit organization (1 response each).

^*Bolded p-values (<.05) indicate significant differences between trained and untrained groups.

Prior to witnessing the most recent overdose, 38% of survey participants (n = 70) were trained in overdose prevention. Among the subset of trained women who were asked about an overdose training venue (n =38), 86.8% indicated they were trained at Prevention Point Philadelphia and the rest were trained somewhere else (data not shown). Trained women were more likely than untrained ones to engage in past 30-day risky drug use, i.e., injecting drugs, using any opioid or heroin, and mixing opioids with benzodiazepines. However, difference in fentanyl use between the two groups was insignificant. Compared to untrained women, trained women had a higher median number of witnessed overdoses, but the two groups did not differ on the median number of lifetime personal overdoses.

Qualitative interviewees had sociodemographic characteristics comparable to those in the quantitative sample. The mean age was 35.4 years and the majority were White (65.8%). Before witnessing the most recent overdose, a higher proportion of the interviewees (66% or 25 women) than survey participants (38%) were trained in overdose prevention. Among trained interviewees with a known place of overdose training (n = 21), 18 or 86% were trained at Prevention Point Philadelphia (data not shown).

Circumstances of the last witnessed overdose

More than three-quarters (77%) of survey participants saw the most recent overdose within the past year. Outdoor places, including streets, alleyways, parks, and abandoned lots, were the most common locations of the last witnessed overdose (62%). Strangers (58%) were the most frequent type of overdose victims. Training status was positively associated with witnessing the last overdose in the past year, at an outdoor place, and reporting strangers as overdose victims (Table 1). Of note, among qualitative interviewees, even a higher proportion witnessed the last overdose at an outdoor place (68.4%) and reported a stranger as an overdose victim (60.5%).

Almost all survey participants (97.8%) indicated that a victim had some combination of typical opioid overdose signs at their last witnessed overdose; the top three included unresponsiveness (65.6%), inability to breathe on her/his own (52.2%), and being unconscious or in a coma (51.6%). The remaining participants (2.2%) encountered other symptoms in victims, such as seizures, stiffening, and foam at the mouth (data not shown).

Overall, making a 911 call was the most prevalent overdose response among survey participants (48.4%), followed by slapping, hitting or pinching the overdose victim (19.9%) and putting water in the victim’s face (18.8%) (Table 2). The other types of recommended response— naloxone use and rescue breathing/CPR — were used by 15.6% and 14.5% of women accordingly. Compared to untrained women, trained women were more likely to administer naloxone (p < .001) and less likely to keep the victim awake (p = .02).

When combinations of recommended actions were taken into account (Table 3), 64% of the quantitative sample used at least one recommended response (either a 911 call, naloxone, or rescue breathing/CPR). Trained women tended to use some recommended response more frequently than untrained (72.9% vs. 58.6%) though the difference was not significant (p = .05). Among 51 trained women who utilized some recommended response, 37 or 72.5% used only one recommended action.

Overall, half of the quantitative sample utilized one recommended action, 13.4% used two actions and only one participant (untrained) used all three recommended actions during the most recently witnessed overdose. Trained participants (20.0%) were more likely than untrained (13.4%) to use two recommended responses (p =.04). Both trained and untrained women combined 911 call with rescue breathing/CPR in approximately equal proportions (7.1% and 9.5%), but only trained women utilized both 911 call and naloxone administration (Table 3). Quantitative and qualitative results regarding each recommended action undertaken by the participants are reported in more detail below.

Recommended response at the last witnessed overdose

911 call

Quantitative results.

A 911 call, potentially mixed with other strategies, was the most frequent overdose response among both trained (48.6%) and untrained (48.3%) participants (Table 2). When a 911 call was used as the only recommended action, it was more prevalent among untrained (37.9%) than trained (31.4%) participants, however the difference between the two groups was insignificant (Table 3). Of note, street involvement characteristics (unstable housing, survival sex, drug use or sales in public places), which could potentially expose women to more frequent interactions with the police and subsequently deter them from making a 911 call, were not significantly associated with this type of response (data not shown).

Qualitative results.

Approximately half of untrained (7 out of 13) and only a third of trained (8 out of 25) interviewees dialed 911 during the last witnessed overdose. Qualitative interviews indicated that a 911 call was often made in the absence of naloxone. When these two recommended actions were applied together, it usually meant that naloxone did not work and a 911 call was made as a last resort.

When my boyfriend overdosed a couple of months ago, the one bottle of Narcan wasn’t enough. He needed three of them to come out of his funk. It was scary. I had to call 911.

(White, 25, untrained)

Interviewees also identified a range of situations and reasons as to why they did not dial 911. On three occasions (2 trained participants and 1 untrained), a 911 call was not made simply because the women did not have a cell phone. However, they approached a police officer for help or asked bystanders to dial 911 for them.

In other cases, however, participants had profound reservations about calling 911. The fear of police related to the penalization of certain practices (particularly, the occupation of abandoned buildings) discouraged women from making a 911 call in a timely fashion. One trained respondent who did not have a habit of carrying naloxone disclosed that her network had an informal rule of avoiding police involvement as much as possible. That practice led to the death of a person whose overdose the participant witnessed most recently:

I saw her [overdose victim] when she was out of there, because they [the group] were trying to get me to help. I poured water on her … obviously it didn’t help, but… yeah, I saw that she had went out… and the paramedics took her… ODing [overdosing] is really common, so they [the group] just assumed that she will come out of it eventually, and then if she don’t… come out of it to their standards, that’s when they call the paramedics… The lives that we’re living, we don’t wanna draw too much attention to what we’re doing… y’know what I mean? So… it’s a bad— it’s a bad way to think, but… if we can avoid calling 911, we do.

(African-American, 26, trained)

Even when respondents made a timely 911 call, fear of police deterred them from another recommended response, such as staying with a victim until the arrival of emergency medical services:

When I saw the ambulance going to the abando [abandoned building], I left. I took off, because I don’t wanna no part of it, you know, and uh, I feel bad, but what else was I supposed to do? I’m not going to jail, I didn’t want the police, the ambulance, nobody asking me questions.

(Mixed race, 43, untrained)

Naloxone administration

Quantitative results.

Training was strongly associated with naloxone use: during the last witnessed overdose, one-third (32.9%) of trained participants and only 5.2% of untrained ones administered naloxone to an overdose victim (Table 2). According to the SAMHSA toolkit (2018), in opioid overdose, naloxone should be administered after making a 911 call. Yet, 14 out of 23 trained women who administered naloxone did not use other recommended strategies (Tables 3 and 2, respectively) and trained women were more likely than untrained ones to use naloxone as the only one recommended overdose response (Table 3). Additionally, only trained participants combined naloxone with a 911 call (Table 3). Still, among the subset of trained women who were asked about the frequency of naloxone carrying (n = 38), only 55.3% said they usually carried naloxone on an everyday basis or several days a week, while more than a quarter (28.9%) reported they never carried naloxone (data not shown).

Qualitative results.

Administering naloxone was a common witnessing behavior of trained women in the qualitative sample: approximately half (12 out of 25) administered their own naloxone, one trained woman provided naloxone for somebody else to use, and two other trained participants administrated somebody else’s naloxone during the last witnessed overdose.

Trained women expressed an enthusiastic attitude toward naloxone (‘I’m really pro-Narcan, so I always tell people to get Narcan,’ White, 23) and noted the simplicity of using the medication (‘It doesn’t take Einstein to do it. It’s nasal spray. I cannot mess it up,’ White, 38). The drug was described as very effective in reversing an overdose of a seemingly ‘dead’ victim, especially in comparison with folk-based strategies unsuccessfully tried by other bystanders:

The girl […] was on the floor…she was like she was dead…people were trying to put water on her, revive her, shake her. I said, “Don’t do that!”. I hit her with Narcan and she came back to her pulse and that was it.

(Pacific Islander, 34, trained)

Every time trained participants administered naloxone it reversed an overdose except for one case when the overdose was likely related to fentanyl, a particularly potent opioid, or substances other than opioids:

I lost somebody… It was a friend - well, an associate of mine. I gave them a Narcan, and it just didn’t work… I don’t know if it was too much or if it was cut with something, because they used to do that to dope [heroin] all the time. They did it, and they went out - stopped breathing, turned blue, everything. I hit them with Narcan three times - nothing. […] The paramedics came. They hit him with Narcan - nothing.

(White, 34, trained)

Access to naloxone was an important benefit of overdose prevention training, however, on its own, training did not always guarantee that naloxone was carried at all times. Ten trained women did not have naloxone on them at the moment of the most recently witnessed overdose for a variety of commonplace reasons, including losing it, using it up, forgetting it at their home, or not having a habit of carrying it. One unstably housed woman (African-American, 26) shared that she usually carries many belongings and always loses everything, including naloxone. Another participant (White, 46), a methadone patient, explained that she only had the injectable form of naloxone (i.e., vials), however her methadone program did not allow carrying syringes, so she kept the medication home. The other two women were trained in settings, such as rehabilitation centers, that did not provide take-home naloxone. As one stated:

They just trained us, they have their own, like, personal Narcan, and they show you how to use it […]. They don’t give it to you.

(African-American, 33, trained)

At the same time, access to naloxone was not limited by participation in overdose prevention training. Two women who were not trained in overdose prevention, but administered naloxone during the last witnessed overdose reported that they knew about the antidote from their prior first aid/CPR training. One of them, a former marine, obtained naloxone through the Veteran’s Administration. The other woman, with a nursing background, acquired naloxone through her personal network. To learn how to use it, she ‘just watched somebody do it in Kensington’ (White, 25). Another untrained woman did not use naloxone at the last witnessed overdose, but seized an opportunity of getting it at a harm reduction site (‘There was the Narcan kit here [at Prevention Point Philadelphia]… and … I took it out,’ White, 46).

Some participants also mentioned the existence of ‘collective’ naloxone in the drug user community that can be obtained from peer users or sellers:

That guy under there has a whole backpack full of Narcan, and he sells drugs too. […] He doesn’t sell the Narcan. He keeps the Narcan for when people overdose under the bridge. It’s like his spot. It’s like this old, Puerto Rican’s spot where he sells drugs. Everything’s down there. Everything’s under there like heroin, Suboxone, Xanax, pills. Everything is down there. […] Yeah. It’s pretty much an under-the-bridge pharmacy. This guy, I don’t know… the cops come down under there all the time, but this guy always gets away. He always has Narcan in one of his book bags. When somebody falls out, he just gives them Narcan like it’s nothing and just leaves them laying there. They come out of their funk almost every time.

(White, 25, untrained)

Notably, both trained and untrained participants mentioned that overdose victims were frequently unhappy with being administered naloxone due to the strong withdrawal effect it produces. An untrained participant had such an unpleasant personal experience with naloxone that she became opposed to the entire idea of overdose prevention training, as well as administering naloxone to other people:

I just remember that feeling … the worst feeling in the world when they put that [naloxone] into you […]. Not that it sent me into withdrawal, it sent me into this… my whole body got cold. It felt like I was freezing to death from the inside out if that can make any sense […] I don’t know. I’d be afraid to hurt someone else.

(White, 47, untrained)

A trained participant stated that she would only administer naloxone in case of an extreme emergency, i.e., total unresponsiveness, since many people who use drugs ‘do not want’ naloxone. She illustrated by describing the most recent experience of witnessed overdose when an overdose victim rejected naloxone offered by an ambulance:

There was like a woman, she was like blue. Like blue-blue-blue… And then they [ambulance] came: “Oh, we’re going to give you the Narcan”… She went like “No-no-no-no. I’m okay, I’m okay, I’m okay.” And just walked away, you know.

(Mixed race, 33, trained)

Rescue breathing/CPR

Quantitative results.

According to the SAMHSA overdose prevention toolkit (2018), supporting a person’s breath is a recommended way of responding to an opioid overdose, which should follow checking for overdose signs, calling 911, and administering naloxone. Moreover, this measure ‘may be lifesaving on its own’ (SAMHSA, 2018, p. 7). Yet, rescue breathing/CPR was the least utilized type of recommended overdose response compared to both 911 call and naloxone use (Table 2). Untrained participants used rescue breathing or CPR more frequently than trained counterparts as the only recommended response (6.1% vs. 1.4%) or in combination with a 911 call (9.5% vs. 7.1%). Still, the difference between the groups was insignificant (Table 3).

Qualitative results.

Only 3 interviewees—2 untrained and 1 trained—provided rescue breaths or CPR to an overdose victim during the last witnessed overdose. In all three cases, rescue breathing preceded or followed a 911 call from the witness or a bystander, but it was never combined with naloxone administration.

In the Rite-Aid parking lot, there was a guy saying, “Help me, help me” and pulling a guy out of a car. He wasn’t breathing. I ran over. There was no ambulance. There were a couple people around calling 911. The guy was just standing there. Nobody was doing anything. I ran over and gave him… I felt his pulse. No pulse at all, so I start CPR on him. Then I got a faint pulse. Then I started the sternum rub. He started [makes coughing sound]. I brought him back.

(White, 42, untrained)

Interviews also revealed that the need to get in close physical contact with an overdose victim may preclude some participants from giving rescue breaths:

The most recent one that I’ve seen was a guy in the park […]. He kind of looked like he was bent over, but he was leaning over like this and I’d seen that his pants were wet and he had a throw-up in his hands. So, I’m like, “Oh my God, what’s wrong with him?”. And I’m looking, I’m like, “Sir, sir!” and he’s not answering me. But I’m not going to go near him only because he had a throw-up. Like I’m not. I know they say you should do the mouth-to-mouth stuff, but I’m not touching someone’s mouth when they have a throw-up - would you do it? No, right? So, I called the ambulance and the ambulance came.

(White, 27, trained)

Discussion

To the best of our knowledge, this is the first study that described response to a witnessed overdose in a group of predominantly street-involved women with a history of illicit drug use—a population at high risk for both personal and witnessed overdose. A key finding of this study is that overdose prevention training improved women’s response behavior by increasing the chances of naloxone administration and using two recommended rescue techniques. At the same time, the study identified structural and situational barriers that interfered with women’s ability to utilize recommended overdose responses.

A significant association between overdose prevention training and naloxone administration was established and corroborated by qualitative data showing that a greater number of trained than untrained women used naloxone at the last witnessed overdose. This finding is consistent with previous research reporting the increased rates of naloxone administration among overdose trainees (Sherman et al., 2008; Strang et al., 2008; Tobin et al., 2009; Wagner et al., 2010). Yet, only a third of all trained women used naloxone at the last witnessed overdose. While naloxone may not be needed in all overdose cases, especially when the victim was revived with stimulation, rescue breathing, or recovered on her/his own (Lankenau et al., 2013), qualitative data pointed to two barriers to consistent naloxone use. First, participants, the vast majority of whom were unstably housed, reported problems with naloxone storage and carrying the antidote on a regular basis. Wagner and colleagues (2010) described a similar impediment to naloxone administration in a comparable sample of overdose responders commonly reporting unstable housing. The other barrier can be described as the unwillingness of drug-using responders to contribute to their peers’ opioid withdrawal by giving them naloxone—a phenomenon also reported in other studies (Bowles & Lankenau, 2019; Lankenau et al., 2013; Strang et al., 2008). Interestingly, not only trained women, but also a small proportion of untrained participants (5.2%) administered naloxone at the last witnessed overdose. The availability of naloxone to untrained people through social networks was documented in prior research (Bowles & Lankenau, 2019; Doe-Simkins et al., 2014). It is likely that certain amounts of naloxone distributed at PPP overdose prevention trainings ‘diffused’ from trained participants to their untrained counterparts (Sherman et al., 2009) or even to drug sellers, especially given that Kensington neighborhood, a predominant location of PPP and present study, is also a large open-air drug market.

The frequency of calling 911 and rescue breathing/CPR was similar among trained and untrained women. This finding fits the results reported by Doe-Simkins and colleagues (2014) who found no significant differences in the utilization of rescue breathing or calling 911 by substance-using overdose rescuers in the pre- and post-training periods. There are several possible explanations for this finding. First, for people who use drugs, getting access to naloxone can be the major advantage of overdose trainings, while they may already have some pre-existing knowledge of other overdose prevention methods (Sherman et al., 2008). Second, since all participants in this study were recruited through a harm reduction program, the untrained group could be exposed to overdose prevention education through channels other than trainings. Those sources can include, for example, interactions with PPP staff, informational materials available at PPP, as well as interactions with trained peers met at PPP. Finally, calling 911 during a medical emergency is a common-sense behavior regardless of training, and rescue breathing skills could be obtained even before training, for example, through a medical background or first aid classes at high school or college.

Almost half of survey participants called 911 as a standalone strategy or in combination with other responses. Moreover, a 911 call was the most prevalent response in the quantitative sample overall, as well as among trained and untrained participants separately. This finding was surprising given that overdose witnesses who actively use substances frequently avoid making a 911 call due to possible contact with the police (Holloway et al., 2018; Lankenau et al., 2013; Sherman et al., 2008). This finding however may be explained by the fact that the majority of overdoses seen by the participants occurred outdoors (62%)—places where overdose witnesses feel more comfortable dialing 911 compared to private locations (Lankenau et al., 2013; Lim et al., 2019; Tracy et al., 2005). Besides fear of police, other barriers to calling 911 included reliance on other methods of overdose reversal, such as naloxone, and the lack of a cell phone. As for the latter, it is often overlooked that possessing a cell phone can be a challenge for street-based people who use drugs. Only 58% of participants of our study owned a cell phone. Qualitative findings also indicated that the simple fact of calling 911 should not be considered a life-saving measure on its own unless the call was made in a timely fashion or a witness stayed with the victim until the arrival of ambulance.

Rescue breathing or CPR was the most underutilized recommended response in the quantitative sample (14.5%), including trained women (11.4%). Rescue breathing requires close physical contact with a victim and witnesses may be reluctant to provide such assistance because a victim may exhibit unpleasant overdose symptoms, for example, vomit contaminates in her/his mouth, as one qualitative interviewee revealed. This is exacerbated by the fact that the majority of overdose victims in this study were strangers who seem to be the least likely candidates for this type of response. We compared responses at the last witnessed overdose among trained participants with results reported by Lankenau and coauthors (2013) who also examined responses of trained drug-injecting people at the last witnessed overdose. Compared to the Lankenau study, trained participants in our study reported a lower frequency of rescue breathing (11% vs. 33%), but a substantially higher prevalence of witnessing an overdose of a stranger (76% vs. 13%). Yet, rescue breathing has a certain advantage over other recommended responses as it formally does not require carrying a device (cell phone or naloxone), though trainees are advised using a disposable face shield rather than providing direct mouth-to-mouth resuscitation.

Notably, trained participants rarely followed overdose prevention guidance of applying a combination of overdose rescue techniques and using them in a particular order (first, a 911 call, then naloxone, and then rescue breathing)—a finding also reported by Lim et al. (2019) and Lankenau et al. (2013). Among trained participants who utilized any recommended response, almost three-quarters employed only one recommended technique. Moreover, in cases when two recommended techniques were applied, the second one was often used as a substitute for the first one that did not work; a typical example was making a 911 call after naloxone did not revive a victim. Qualitative data showed that women resorted to naloxone administration as an effective, fast, and low-risk strategy compared to the other equally effective, but time-consuming or riskier responses (for example, supplementing naloxone use with a 911 call or making a 911 call and waiting for help to arrive). Furthermore, our study showed that rescue breathing, apparently perceived as a less effective method, was more deliberately combined with a 911 call, than naloxone use. Future studies should investigate how the perceptions of effectiveness and risk impact the sequence of overdose responses among street-involved responders.

Women’s street involvement significantly affected circumstances of witnessed overdose and type of overdose response. Quantitative findings indicated that most overdoses reported in this study were experienced by strangers, witnessed in public settings, and addressed with a 911 call rather than rescue breathing or CPR. In contrast, studies that examined overdose witnesses who used drugs but were not predominantly street-involved reported a higher share of overdoses seen in private places, experienced by friends or partners/family, and handled with the use of rescue breathing (Doe-Simkins et al., 2014; Lankenau et al., 2013). Moreover, qualitative findings revealed the adverse effect of street involvement on all three types of the recommended response. Due to unstable housing, women were less prone to carry naloxone or possess a phone, i.e., make a 911 call. The absence of permanent housing also contributed to the participants’ use of abandoned buildings, an illegal activity, thus delaying a 911 call related to an overdose emergency. Finally, all street involvement characteristics increased women’s odds of seeing an overdose of a stranger, that discouraged them from giving rescue breaths or staying with the victim.

Our research also raises an important question about the potential effect of gender on the level and nature of overdose responses reported in this analysis. Research grounded in the social role framework postulates that prosocial behaviors tend to align with traditional gender roles, so that women are more likely to extend care in the context of close relationships, while men are more likely to provide heroic emergency helping, especially directed at strangers and in unfamiliar physical settings (Diekman & Clark, 2015; Eagly, 2009). Since a majority of witnessed overdoses reported in this study were experienced by strangers in public locations, we can assume that based on the social role theory, the overall rate of any overdose response could have been higher in a sample that included men. At the same time, since women enact a higher level of evidence-based overdose prevention behaviors than men, such as naloxone possession (Tobin et al., 2018) or making a 911 call (Ambrose et al., 2016; Lim et al., 2019), it is possible that the share of recommended responses relative to any overdose response is higher in our data than it would have been among men-only or mixed-gender samples. Of note, gender can potentially influence overdose responses in other ways: men may be more prone to help to women than men (Eagly, 2009) and the mere presence of women bystanders during an overdose may increase the chances of a 911 call (Tobin et al., 2005). Future studies need to provide more insights into power dynamics in mixed gender groups of overdose bystanders and how the responder’s gender and relationship to the victim affects overdose response.

This study has important implications for future overdose prevention efforts. Our findings indicated that improved access to naloxone constitutes a major benefit of overdose training for women lay responders. Hence, training programs should put more effort into explaining the life-saving potential of naloxone, even despite its potential withdrawal effect, and identifying strategies to address possible anger or aggression of overdose victims experiencing withdrawal symptom upon revival. Furthermore, in light of the increasing share of fentanyl-involved overdose deaths in the U.S. (Wilson et al., 2020) and given fentanyl’s high potency and rapid onset (Comer & Cahill, 2019), lay responders should be trained in multiple naloxone administrations and provided with a greater number of naloxone doses to avert the risk of inadequate overdose reversal (Moss & Carlo, 2019). Training curriculum should also emphasize a value of proper sequence of overdose prevention techniques and give sufficient attention to teaching and practicing rescue breathing.

Our study also reiterates the importance of multiple community-based distribution points that may be particularly beneficial for unstably housed responders who are unable to carry naloxone on a constant basis. In fact, as our data collection was close to completion, the city of Philadelphia began wider naloxone distribution through various community sites, including nonprofits, pharmacies, and healthcare facilities (City of Philadelphia, 2019). Additionally, our findings highlight the utility of distributing naloxone via peers who use drugs or small-scale dealers. Interestingly, research has found that secondary naloxone distribution, i.e., allowing pickup for multiple naloxone kits and distributing via social networks, can be as effective in reducing overdose mortality as increasing the number of community pickup sites from one to ten (Keane et al., 2018).

This study has several limitations. First, the study used cross-sectional data, so the results do not indicate causality. At the same, since the independent variable (overdose prevention training before the last witnessed overdose) was chosen to precede the dependent variable (response at the last witnessed overdose), a temporal order can be assumed and training may be treated as predicting a certain type of overdose response. Second, our analysis of recommended responses was limited to the three actions, such as a 911 call, naloxone administration, and rescue breathing or CPR. We did not examine the other types of recommended response, such as checking overdose signs, including stimulation with a sternal rub (though it surfaced as an open-ended answer), or staying with the victim until help arrived. Therefore, our results should be treated with an understanding that the above-mentioned types of recommended response were not taken into account. Third, all data were self-reports and subject to recall bias and desirability bias. Fourth, Kensington area of Philadelphia has an open drug scene and the majority of the participants witnessed outdoor-based overdoses experienced by strangers. It may affect the external validity of quantitative results and transferability of qualitative results since other settings may have different combinations of overdose places and victim types than Philadelphia. Finally, we did not quantify factors that can contribute to an overdose response in addition to training, which was beyond the scope of this study. Nonetheless, thanks to the qualitative results, we can outline those factors (most importantly, overdose place, victim type, and the presence of other bystanders) and examine their influence on overdose response in future research.

In conclusion, this study established a positive effect of overdose prevention training on witnessing behavior of predominantly street-based women who use illicit drugs. Compared to untrained women, their trained counterparts were more likely to administer naloxone and provide two types of recommended assistance at the last witnessed overdose. At the same time, we identified a number of structural and situational barriers to the consistent use of recommended responses, including the avoidance of police, street involvement, as well as the predominance of outdoor-based overdoses and strangers as typical overdose victims. Our study contributes to the growing body of research aiming to provide scientific evidence to address the broader opioid overdose crisis. Findings indicate that street-based women who use drugs, a marginalized population that is vulnerable to both personal and witnessed overdose, can save the lives of peers with proper training and support.