Drug-related harm coinciding with income assistance payments: results from a community-based cohort of people who use drugs

Abstract

Background and Aims

Income assistance is critical to the health and well-being of socioeconomically marginalized people who use illicit drugs (PWUD). However, past literature paradoxically identifies unintended increases in drug-related harm coinciding with synchronized payments that may magnify signals for drug use. The scope of such harm has not been fully characterized among non-institutionalized populations. This study examined sociodemographic, health and drug use-related correlates of payment-coincident drug-related harm.

Design

This observational study using data from prospective community-based longitudinal cohorts of PWUD between December 2013 and May 2018.

Setting

Vancouver, British Columbia, Canada.

Participants

A total of 1604 PWUD receiving monthly income assistance. Our sample included 586 (36.5%) women, 861 (53.7%) non-white participants, and 685 (42.7%) people living with HIV.

Measurements

Primary outcome was a self-reported composite measure of drug-related harm in the past 6 months coinciding with income assistance, including higher frequency substance use, non-fatal overdose, and service barriers or interruptions. Subanalyses disaggregated this outcome.

Findings

Payment-coincident drug-related harm was reported among 77.7% of participants during the study period. In multivariable models, key correlates positively and significantly associated with payment-coincident harm included: street-based income generation (Adjusted Odds Ratio [AOR]=1.48, 95% Confidence Interval [CI]=1.26–1.74, P<0.001), sex work (AOR=1.66, 95% CI=1.35–2.04, P<0.001), illegal income generation (AOR=1.57, 95% CI=1.35–1.83 P<0.001); homelessness (AOR=1.34, 95% CI=1.13–1.58, P<0.001); exposure to violence (AOR=1.31, 95% CI=1.03–1.66, P=0.032), daily crack-cocaine use (AOR=1.99, 95% CI=1.59–2.50, P<0.001); heavy alcohol use (AOR=1.64, 95% CI=1.37–1.97, P<0.0001) and injection drug use (AOR=2.55, 95% CI=2.01–3.23, P<0.001). In subanalyses, specific harms were more likely among individuals reporting social, structural and health vulnerabilities.

Conclusions

In Vancouver, Canada, people who use illicit drugs who receive income assistance report high prevalence of payment-coincident drug-related harm, particularly people experiencing socioeconomic and structural marginalization or engaging in high-intensity drug use.

Introduction

Cash transfer benefits provide financial support and are considered a crucial way that states mitigate extreme poverty (1, 2), improving health and reducing mortality among recipients (3, 4). For people who use drugs (PWUD), income assistance is additionally associated with reduced recidivism (5) and substance use disorder (SUD) treatment retention (6). Notably, increased spending commonly follows income payments (7), including income assistance, which in North America are predominantly distributed on a synchronized monthly basis (11, 12). For PWUD, coordinated benefit receipt may unintentionally initiate intensified drug use (8), particularly in areas with concentrations of recipients that may magnify social signals for use (9, 10), inadvertently driving substance use-related harm (13). This phenomenon has important implications for the support of marginalized populations at elevated risk for suboptimal health, social, and economic outcomes.

Despite the considerable benefits of income assistance for socioeconomically marginalized PWUD, the substantial increases in drug-related harm paradoxically linked to synchronized income assistance disbursement by past research includes wide-ranging morbidity, mortality, and barriers to services. Harms include increases in drug use (14–16); non-fatal overdose, fatal overdose and drug- and alcohol-associated deaths (15, 17–22); emergency department admissions and hospitalization (14, 15, 23–27); ambulance service calls (21); leaving hospital against medical advice (18, 28, 29); sobering or detoxification unit admissions (21, 30); substance use and HIV treatment interruption (28, 29, 31); public disorder and police activity (21, 23, 32); 911 service calls (23); and service access barriers (33). The unintended impacts of coordinated payments for PWUD, service provision and street-based disorder are widespread and severe. Despite early claims that assistance benefits prompt increases in drug use, analyses have found that synchronized monthly disbursement shapes timing and intensity of use rather than overall levels of use. (16, 34–36).

Nearly all prior research examining the “cheque effect” relies on institutional, coroner’s, health service or other administrative data, restricting study populations and available indicators, with extremely limited data measuring drug use, drug use timing, and intoxication (9, 13, 36). Limited observational studies include treatment-enrolled individuals (16, 36), with little information about the prevalence of drug-use and drug-related harm in community-based settings. Consequently, there is significant potential for misestimating the burden of payment-coincident drug-related harm. Further, using administrative data from institutional settings may overlook the scope of harms associated with payments. Large influxes of cash in communities with concentrations of assistance recipients may substantially impact drug scene activity in terms of drug dealing; drug debt repayment; informal, prohibited and illegal income generation (e.g., sex work or theft); and associated public disorder and violence (37, 38). Previous research may therefore preclude a robust understanding of the true so-called “cheque effect.”

Therefore, we undertook analyses of the prevalence, correlates and scope of drug-related harm coinciding with income assistance payments among community-recruited, prospective cohorts of PWUD in Vancouver, Canada, where cash transfers occur once per month, generally on the last Wednesday of the month. While there is longstanding understanding among community members, service providers and emergency responders that drug-related activity increases substantially around benefit disbursement (18, 21), there is little understanding of the prevalence and scope of payment-coincident drug-related harm in community-based populations. Our specific research questions are:

1. Among government income assistance recipients who use illicit drugs, what is the prevalence of drug-related harm coinciding with monthly payments?

2. What are the socio-demographic, socio-economic and drug-related correlates of harm coinciding with monthly assistance payments among PWUD?

3. Are there characteristics, exposures or activities associated with specific types of payment-coincident drug-related harm?

Methods

Design

We analysed data from the Vancouver Injection Drug Users Study (VIDUS) and AIDS Care Cohort to evaluate Exposure to Survival Services (ACCESS), two ongoing prospective cohort studies of HIV-seronegative (VIDUS) or HIV-seropositive (ACCESS) PWUD. Detailed previously (39, 40), the studies began enrolment through self-referral and community-based methods in May 1996. Participants complete baseline and semi-annual follow up study visits that include an interviewer-administered questionnaire and the provision of a blood sample for serologic testing. The current analysis used data from all VIDUS and ACCESS baseline and semi-annual follow from December 2013 through May 2018, during which increased community and scientific interest in the topic prompted the inclusion of questions specific to drug-related harm around assistance payments in the questionnaire. Participants receive an honorarium of $40 CAD at each study visit. Both cohorts received ethical approval from the Providence Health Care/University of British Columbia Research Ethics Board and adhere to STROBE guidelines for survey-based longitudinal studies (41). The analysis was not pre-registered on a publicly available platform and results should be considered exploratory.

Study sample

Individuals are eligible for the cohorts if they injected (VIDUS) or used (ACCESS) illicit drugs other than cannabis in the previous month, live in Greater Vancouver at enrolment, and provide written informed consent. This study restricted the sample to individuals receiving monthly government income support. The study used participants’ first available observation in which data on payment-coincident drug-related harm was available (December 2013 or later) as the analytic baseline.

Measures

Outcome

Reports of harm were derived from the question: “Did any of the following things ever happen to you in the days around cheque day?” to capture increased drug-market activity directly before and subsequent to payments. Response options included: used drugs or drank more than usual, had a non-fatal overdose, was forced to settle a drug debt, could not access a health or social service, was unable to use InSite (a supervised injection facility), had contact with police, left a hospital stay, left hospital against medical advice, left SUD treatment, or other. An affirmative answer to any response option was considered an incident of payment-coincident harm. Participants specified whether they had ever experienced the harm in question, and whether that harm occurred in the six months prior to interview, allowing for binary outcome measures incorporating (1) lifetime and (2) past six month time frames.

Covariates. We selected a priori covariates based on previous theoretical and empirical research on payment-coincident harm (13, 17, 18, 29, 32). To account for differences in socio-demographic characteristics, we considered age (per year older), gender (female vs. male), white ethnicity (yes vs. no) and educational attainment (less than high school vs. high school or greater). Given extensive research documenting the relationship between social-structural disadvantage and drug-related harm (37, 42), we additionally considered indicators of social-structural marginalization reported in the six months prior to interview. These included binary indicators of street-based income generation, including informal recycling, car window washing, or panhandling; sex work; and illegal activities such as drug dealing, theft or other crime; homelessness (yes vs. no); recent incarceration, defined as having been in detention, prison, or jail or penitentiary overnight or longer (yes vs. no); and experiencing physical or sexual violence (yes vs. no). We additionally included residency in the Downtown Eastside (DTES) neighbourhood of Vancouver, which is commonly characterized by an open drug market and economic disadvantage (43). We further accounted for the potential role of health disadvantage given payment-coincident increases in demands for health services (13, 21, 23, 25, 33). Health-related indicators included enrolment in medication assisted treatment for opioid use disorder (MAT; i.e., methadone or buprenorphine-naloxone), enrolment in non-MAT SUD treatment (e.g., withdrawal management, twelve step programs) and the interruption of methadone-based MAT. We also included time updated binary measures of HIV or HCV seropositivity, and ever having been diagnosed with a mental illness. Substance use-related variables referring to the six months prior to interview included binary indicators of daily or more frequent heroin, cocaine, crystal methamphetamine or crack cocaine use. We further considered binary, past six months measures of heavy alcohol use (> four drinks/day on average vs. ≤ four drinks/day) and injection drug use. Finally, we accounted for time trends in the data by controlling for the year of study visit.

Analysis

We first describe the study sample, stratifying descriptive statistics by having experienced payment-coincident drug-related harm in the six months prior to baseline and testing for differences using Pearson’s chi-squared test for dichotomous variables and Mann-Whitney test for continuous variables. To address our first research question, we examined the period prevalence of payment-coincident drug-related harm by assessing the percentage of unique individuals in the study sample reporting our composite primary outcome and its constituent components for each follow up and the overall study period. To answer our second research question, we estimated bivariable and multivariable relationships between our covariates of interest and payment-coincident drug-related harm in the past six months, using logistic generalized linear mixed-effects modelling (GLMM). This approach accounts for repeated measures, allowing for the modelling of correlated and non-normally distributed data (44). We assessed question- and observation-level missing data, and addressed missing data by using standard multiple imputation procedures for longitudinal data, imputing five new data sets and averaging estimates using AMELIA II (45). To answer our third research question, we examined whether key indicators of social, structural and health disadvantage (e.g., mental illness, homelessness, etc.) were associated with specific types of drug-related harm. Sub-analyses used the drug-related harm measure that corresponded to the time frame of the indicator in question (i.e., ever for time-updated lifetime measures; past six months for time-varying indicators). We tested for significance using Pearson’s χ² test or Fischer’s exact test where cell counts are less than or equal to five. All statistical analyses were performed using SAS software version 9.4 (SAS, Cary, NC). All p-values are two-sided.

Results

Study Sample

From December 2013 to May 2018, 1604 participants receiving provincial government income assistance completed a median 6 of 8 possible study visits during the study period (interquartile range [IQR] 3–8 visits). There were 1393 (13.7%) missed observations and variable-level non response ranged from 0·01% (>daily crack-cocaine use) to 2·5% (educational attainment) among available observations. When defined as having the last study visit more than two years before the end of the study period, 282 participants (17·6% of the sample) was lost to follow up. There was no significant difference in the likelihood of our outcome between those that were and were not lost to follow up (p=0·366).

Participants’ median age at baseline was 46.6 years (IQR: 37·8–53·3), 586 (36·5%) individuals self-identified as women, 731 (45·6%) as white, and 795 (49·6%) had less than high school education. Table 1 shows baseline sample characteristics. Individuals who experienced harm during the study period were more likely to be younger, women, reside in the DTES, exposed to violence, engaged in sex work, street based or illegal income generation, have had a mental health diagnosis, or have interrupted their MAT treatment in the six months prior to baseline. They were also more likely to use heroin, cocaine, methamphetamine or crack-cocaine on a daily or more frequent basis, have injected drugs or engaged in heavy alcohol use in the six months prior to baseline (all p<0·05).

Table 1·.

Baseline characteristics of people who use illicit drugs in receipt of government income assistance in Vancouver, Canada, stratified by experiencing drug-related harm coinciding with payment disbursement at any point during the study period (2013–2018)

Characteristic	Total (%) (n = 1604)	Harm coinciding with income assistance payments at baseline		p - value
		Yes (%) (n = 1262)	No (%) (n = 342)
Socio-demographic
Age (med, IQRa)	46·6 (37·8–53·3)	46·2 (37·6–52·8)	48·7 (38·7–55·1)	0·002
Female gender	586 (36·5)	477 (38·3)	109 (30·4)	0·007
White ethnicity	731 (45·6)	570 (45·7)	161 (45·0)	0·809
≥ High school education	795 (49·6)	625 (50·2)	170 (47·5)	0·354
Social-structural vulnerability
Homelessness b	358 (22·3)	286 (23·0)	72 (20·1)	0·236
DTES Residency bc	970 (60·5)	785 (63·0)	185 (51·7)	<0·001
Incarceration b	117 (7·3)	98 (7·9)	19 (5·3)	0·105
Exposure to violence b	242 (15·1)	205 (16·5)	37 (10·3)	0·004
Street-based income bd	338 (21·1)	293 (23·5)	45 (12·6)	<0·001
Sex work b	203 (12·7)	185 (14·8)	18 (5·0)	<0·001
Illegal income be	414 (25·8)	368 (29·5)	46 (12·8)	<0·001
Health & Health Service Use
MAT enrollment bf	844 (52·6)	669 (53·7)	175 (48·9)	0·096
Non-MAT SUD treatment b	169 (10·5)	129 (10·4)	40 (11·2)	0·670
MAT treatment interruptionb	337 (21·0)	290 (23·3)	47 (13·1)	<0·001
Mental health diagnosis g	1041 (64·9)	835 (67·0)	206 (57·5)	<0·001
HCV positive g	1293 (80·6)	1024 (82·2)	269 (75·1)	0·081
HIV positive g	685 (42·7)	540 (43·3)	145 (40·5)	0·339
Drug use
Daily heroin use b	379 (23·6)	318 (25·5)	61 (17·0)	<0·001
Daily cocaine use b	84 (5·2)	73 (5·9)	11 (3·1)	0·037
Daily crystal methamphetamine use b	218 (13·6)	186 (14·9)	32 (8·9)	0·004
Daily crack smoking b	210 (13·1)	197 (15·8)	13 (3·6)	<0·001
Use of injection drugs b	1095 (68·3)	934 (75·0)	161 (45·0)	<0·001
Heavy alcohol use bh	239 (14·9)	205 (16·5)	34 (9·5)	0·001

^aIQR: interquartile range

^bDenotes activities and exposures in the previous 6 months

^cDTES: Downtown Eastside

^dIncludes informal recycling or “binning”, car window washing or “squeegeeing” and panhandling

^eIncludes drug dealing, theft and other criminal activity

^fMAT: medication assisted treatment, i·e·, methadone, suboxone

^gDenotes lifetime history

^hMore than 4 drinks in a single sitting

Prevalence estimates

Payment-coincident drug-related harm was reported in 4179 (47.5%) of 8794 total observations. The period prevalence of harm for each six month follow up period ranged from 41·0% (95% Confidence Interval [CI] 37·9% - 44·1%) to 53·1% (95% CI 49·4–56·3%). Overall period prevalence for the whole study period was 47·5% (95% CI: 46·5–48·6%), with payment-coincident harm reported by 1246 unique individuals for an overall period prevalence of 77·7% (95% CI: 75·6–79·7%). The period prevalence for the components of the composite outcome measure are included in Table 2. Using drugs or drinking more than usual were the most common form of harm, followed by being forced to settle a drug debt, having contact with police, being unable to access a health or social service or supervised injection facility, and experiencing non-fatal overdose.

Table 2·.

Prevalence of drug related harm coinciding with income assistance payments in the six months prior to interview (2013–2018)

Type of drug-related harm	Number of times reported (n, %)a	Overall period prevalence (n, %)b
Used drugs more than usual	3403 (38·7)	1086 (67·7)
Drank more than usual	1180 (13·4)	520 (32·4)
Was forced to settle a drug debt	886 (10·1)	486 (30·9)
Had contact with the police	403 (4·6)	295 (18·4)
Was unable to access a health or social service	348 (4·0)	270 (16·8)
Was unable to access the supervised injection facility	369 (4·2)	253 (15·8)
Experienced non-fatal overdose	212 (2·4)	153 (9·5)
Left a hospital stay	114 (1·3)	101 (6·3)
Left hospital against medical advice	106 (1·2)	90 (5·6)
Left substance use disorder treatment	58 (0·7)	53 (3·3)
Other	18 (0·2)	18 (1·1)
Total	4179 (47·5)	1505 (77·7)

^aNumber of reports and percentage of total observations (n=8794) in which a harm was reported

^bNumber of unique individuals and percentage of the sample reporting a specific harm over the study period.

Correlates of payment-coincident drug-related harm

Table 3 displays GLMM analyses examining correlates of payment-coincident harm. In final adjusted analyses, age and year of interview were inversely associated with drug-related harm. Health, socio-demographic and socio-economic covariates positively and significantly associated with payment-coincident drug-related harm interview include lifetime mental illness diagnosis; homelessness; Downtown Eastside residency; exposure to violence; engagement in street-based income generation, sex work and illegal income generation. Drug use variables significantly and positively associated with the outcome include MAT discontinuation; daily or more frequent crack cocaine use, heavy alcohol use, and injection drug use.

Table 3.

Factors associated with drug-related harm coinciding with government income assistance payments in the past six months among income assistance recipients who use illicit drugs in Vancouver, Canada, 2013–2018 (n=1604)^a

Characteristic	Unadjusted Odds Ratio (95% CI)b	p - value	Adjusted Odds Ratio	p - value (95% CI)
	Socio-demographic
	Age (per additional year)		0·97 (0·96 – 0·98)		<0·001	0·99 (0·98 – 1·00)	0·030
Female gender	1·38 (1·15 – 1·66)	<0·001	1·03 (0·87 – 1·23)	0·736
White ethnicity	1·04 (0·87 – 1·23)	0·688	1·02 (0·87 – 1·20)	0·803
< High school education	1·20 (0·97 – 1·49)	0·065	1·05 (0·88 – 1·25)	0·592
Social-structural vulnerability
Homelessness c	1·97 (1·62 – 2·40)	<0·001	1·34 (1·13 – 1·58)	<0·001
DTES d Residency c	1·56 (1·34 – 1·81)	<0·001	1·22 (1·06 – 1·40)	0·005
Incarceration c	1·93 (1·42 – 2·62)	<0·001	1·23 (0·88 – 1·71)	0·214
Exposure to violence c	1·80 (1·38 – 2·35)	<0·001	1·31 (1·03 – 1·66)	0·032
Street-based income ce	1·79 (1·53 – 2·09)	<0·001	1·48 (1·26 – 1·74)	<0·001
Sex work c	2·45 (2·02 – 2·98)	<0·001	1·66 (1·35 – 2·04)	<0·001
Illegal income cf	2·46 (2·10 – 2·89)	<0·001	1·57 (1·35 – 1·83)	<0·001
Health status & addiction treatment
MAT enrollment cf	1·24 (1·06 – 1·44)	0·007	1·07 (0·93 – 1·24)	0·322
Non-MAT SUD treatment c	0·88 (0·71 – 1·09)	0·221	0·99 (0·81 – 1·21)	0·926
MAT treatment interruption c	1·68 (1·41 – 2·01)	<0·001	1·31 (1·06 – 1·61)	0·014
Mental health diagnosis g	1·32 (1·09 – 1·59)	0·004	1·24 (1·02 – 1·50)	0·030
HCV Seropositivity g	0·99 (0·78 – 1·24)	0·903	0·92 (0·73 – 1·15)	0·440
HIV Seropositivity g	0·90 (0·76 – 1·07)	0·245	1·16 (0·99 – 1·36)	0·074
Drug use
Daily heroin use b	1·92 (1·60 – 2·32)	<0·001	1·09 (0·90 – 1·32)	0·370
Daily cocaine use b	1·58 (1·25 – 1·99)	<0·001	1·24 (0·97 – 1·58)	0·086
Daily crystal methamphetamine use b	1·74 (1·43 – 2·12)	<0·001	1·15 (0·95 – 1·40)	0·143
Daily crack smoking b	2·36 (1·82 – 3·07)	<0·001	1·99 (1·59 – 2·50)	<0·001
Injection drug useb	3·53 (3·03 – 4·11)	<0·001	2·55 (2·01 – 3·23)	<0·001
Heavy alcohol use bh	1·53 (1·26 – 1·87)	<0·001	1·64 (1·37 – 1·97)	<0·001
Year of Interview	0·88 (0·85 – 0·92)	<0·001	0·91 (0·87 – 0·95)	<0·001

^aAnalyses incorporate 8794 total observations

^bCI: confidence interval

^cDenotes activities in the previous 6 months

^dDTES: Downtown Eastside

^eIncludes informal recycling “binning”, car window washing “squeegeeing” and panhandling

^fIncludes drug dealing, theft and other criminal activity

^fMAT: medication assisted treatment, i·e·, methadone, suboxone

^gDenotes lifetime history

^h> 4 drinks in a single sitting

Social, structural and health disadvantage

Figure 1 displays the statistical significance of bivariable associations between subtypes of drug-related harm and social, structural and health disadvantage measures. Subtypes of harm were disproportionately reported by individuals reporting involvement in informal, prohibited or illegal income generation, homelessness, recent incarceration and high intensity drug use. Accidental overdose, being forced to settle a drug debt and encountering barriers to accessing a supervised injection facility were the harms most consistently associated with indicators of disadvantage.

Figure 1.

Associations between specific drug-related harms coinciding with income assistance disbursement and individual social, structural and health exposures among people who use drugs in Vancouver, Canada (2013–2018; n=1604)

Discussion

Governments commonly employ cash transfer benefits to reduce socio-economic and health inequities (48, 49), with important benefits for PWUD (5, 6). Previous literature based on administrative data paradoxically identifies payment-coincident elevations in drug-related morbidity, mortality and service utilization pointing to a so-called “cheque effect” linked to monthly synchronized income assistance disbursement (13, 18, 21, 25). This study extends this work to examine the prevalence, correlates and scope of payment-coincident drug-related harm among community-based cohorts of PWUD. We find an extremely high prevalence of harm and systematic relationships between such harms and social, socio-economic, health and structural disadvantage. In light of previous research identifying considerable harm from income assistance retrenchment among PWUD (5, 48, 49), these results do not imply that denying access to income assistance for PWUD would effectively address these harms. Instead, results identify characteristics and exposures associated with payment-coincident harm that point to potential reforms that could better ensure income assistance meets the needs of PWUD and mitigates unintentionally caused harms.

Analyses to address our first research question concerning the prevalence of payment-coincident drug-related harm revealed that more than three-quarters of participants experienced this outcome during the study. While the most common type of drug-related harm was intensified drug use or drinking, the prevalence of more serious harms, such as non-fatal overdose, being forced to settle a drug debt, contact with police, and challenges accessing services were alarmingly high. Analyses of administrative data that use service indicators, such as leaving hospital against medical advice, find increases in service demands around payments ranging from 9–38% (15, 18, 20, 21, 24–26, 28–30, 33). The difference between our prevalence estimates and those in past research suggest that institutional data sources may underestimate the true prevalence of payment-coincident harm (50, 51). The current study thus advances this literature by using longitudinal data from community-based data to suggest that the scope and scale of harm from synchronized payments may be higher than previously documented.

Analyses to address our second research question about correlates of harm identified systematic relationships with high intensity and high-risk drug and alcohol use, income generation practices associated with socio-economic marginalization, and structural disadvantage. That payment-coincident harm was more likely to be reported among PWUD engaged in high-intensity crack-cocaine, heavy alcohol or injection use parallels previous documentation linking increased frequency or risk of use with elevated morbidity and mortality (52, 53). An influx of cash from income assistance payments is both an individual-level signal for “consumption” akin to that in the general population (7) which may be amplified through socially derived signals for increased use (10). Such social signals may be particularly strong among those engaged in high intensity use, as past research outlines how high risk drug use patterns aggregate in social networks of PWUD (54). These results suggest that efforts to mitigate the harms associated with payments could most effectively support populations engaged in high-intensity and high-risk use, should not overlook alcohol, and need to expand beyond opioid-specific interventions, particularly given the paucity of pharmacological treatments for stimulant use disorders (57, 58).

Multivariate results also indicate that those involved in informal, prohibited and illegal income generating activity are more likely to report payment-coincident harm. This study contributes to growing evidence demonstrating that some types of income generation expose PWUD to social, economic, and physical environments that increase health-related harms (37, 59–62). Individuals engaged in these types of income generation are often supplementing government benefits and may therefore meet basic needs with income from non-benefit sources (61). With access to other income sources, they may therefore spend assistance payments in proximity to payment days (7) in ways more likely to expose them to drug-related harm. PWUD also commonly undertake these activities outdoors, potentially increasing exposure to drug use scenes associated with health harms and service access barriers (64, 65). Our results suggest a central role for activities spatially linked to socio-economic marginalization in shaping vulnerability to harm.

Other results also have implications for spatial dimensions of drug related harm. Individuals who are homeless may be implicated in street-based activity that increases around payment days. Being forced to settle a drug debt or police confrontation, for example, would be more difficult to avoid without shelter. Our findings are consistent with a robust literature documenting the health, social and economic costs of homelessness (57, 64–66). Additionally, residency in the Downtown Eastside was unsurprisingly significant in final analyses. “Cheque Day” or “Welfare Wednesday” is commonly considered a spatially delimited and socio-cultural phenomenon specific to the neighbourhood (9). While our results indicate that drug-related harm around payments may be more likely for DTES residents, it is not unique to such areas as indicated by a recent Rhode Island study that did not find a significant relationship between spatial concentrations of income assistance receipt and increases in fatal overdose coinciding with payments (22).

Considerations related to socioeconomic marginalization and spatial vulnerabilities may also be relevant to findings linking exposure to violence and drug-related harm coinciding with assistance payments. Past research has connected vulnerability to violence to socioeconomic marginalization (37), involvement in illicit drug markets (67) or within-household conflict over resources (68). Our results are consistent with these previous studies. However, there remains a paucity of research in this area, and studies using administrative data may be prone to misunderstanding this relationship given the challenges of measuring the prevalence and scope of exposure to violence (69). Reporting biases implicit in administrative data may result in the significant underestimation of violence (51).

Finally, analyses addressing our third research question about specific types of harm identify the linkages of harms to social-structural disadvantage. Analyses suggest that those exposed to high intensity drug use; informal, prohibited or illegal income; homelessness; or recent incarceration or probation are consistently more likely to experience a broader range of drug-related harm coinciding with payments. Given linkages between these exposures, involvement in drug use scenes and health, social and structural disadvantages (70–72), these findings are unsurprising if alarming.

This study signals a need for strategies that could mitigate these harms. Programmatic or structural changes, such as desynchronizing or changing the frequency of payments, could reduce the burden of harm for PWUD by dispersing social signals for use (7, 10) as well as concentrated service demands for service providers and first responders (73), beneficially impacting payment linked cycles of harm. However, such changes may produce unanticipated consequences by disrupting the predictability of money flows and should be approached with caution (74). Increasing income assistance rates, which were stagnant in the study context during the study period, could reduce the need to generate income through activities these and other analyses link to increased harm (37, 65, 77). Additionally, socio-economic integration through models that support involvement in economic activity could facilitate transitions away from the income assistance system though this may not be feasible or desirable for all recipients (78). Such reforms could reinforce and optimize the critical but improvable role the income assistance plays in reducing poverty and improving health among PWUD (2–6).

This study has limitations. The study population is a non-random sample and findings may not be generalizable to other populations or contexts. Nevertheless, previous studies suggest that the VIDUS and ACCESS samples are reflective of the inner-city Vancouver drug using population (79) and payment linked escalations in drug use and related harm has been observed in other contexts (13). Additionally, data used in the study are self-reported and subject to potential response biases such as recall and social desirability bias. However, the longstanding nature of these studies and experienced technical research staff minimize these threats. Further, response biases are more likely to result in the underreporting of harm and production of conservative estimates. We do not compare payment-coincident harms with those during non-payment periods, and these results are not evidence that harms are more frequent or serious than non-payment periods. Finally, our measures of drug-related harm may not capture the full spectrum of harm experienced by participants and analyses may suffer from unmeasured confounding from variables or interactions not included in analyses.

This study identifies the prevalence and correlates of drug-related harm coinciding with income assistance disbursement using data from longitudinal, community-based cohorts of PWUD. In identifying high prevalence alongside significant linkages with high intensity drug use, structural marginalization and informal and illegal income generation practices that are themselves associated with considerable health harm, this study expands previous understandings of the scope and scale of drug-related harm borne by PWUD around assistance payments. This research speaks to intersections of substance use disorders, socioeconomic marginalization and the physical and social spaces of drug use, as well as how institutional decisions related to cash transfer benefits can have serious consequences. Considerations of whether to manage the impacts of synchronized once-monthly disbursement through support services, or to change the disbursement schedule entirely could have significant implications for the well being of PWUD. Specifically, approaches to reform should interrogate whether policy or programmatic actions can effectively address health equity and safety issues among individuals who face already considerable disadvantage. The cheque effect phenomenon is one where policy structures outside health exert considerable influence over the health of disadvantaged populations, and represents an area where feasible action can be taken to collaborate across areas of substantive jurisdiction to improve health outcomes for people who use illicit drugs.