UBC Faculty Research and Publications

Illicit drug overdose deaths resulting from income assistance payments : analysis of the ‘check effect’… Otterstatter, Michael C.; Amlani, Ashraf; Guan, Tianxiu Hugh; Richardson, Lindsey; Buxton, Jane Jul 31, 2016

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


52383-Otterstatter_M_et_al_Illicit_drug_overdose_deaths.pdf [ 406.02kB ]
JSON: 52383-1.0347526.json
JSON-LD: 52383-1.0347526-ld.json
RDF/XML (Pretty): 52383-1.0347526-rdf.xml
RDF/JSON: 52383-1.0347526-rdf.json
Turtle: 52383-1.0347526-turtle.txt
N-Triples: 52383-1.0347526-rdf-ntriples.txt
Original Record: 52383-1.0347526-source.json
Full Text

Full Text

Illicit drug overdose deaths resulting from income assistance payments: Analysis of the ‘check effect’ using daily mortality data  Authors: Michael C. Otterstattera,b*, Ashraf Amlania, Tianxiu Hugh Guana,1, Lindsey Richardsonc,d, Jane A. Buxtona,b   a BC Centre for Disease Control, 655 West 12th Avenue, Vancouver, British Columbia, Canada, V5Z 4R4  b School of Population and Public Health, University of British Columbia, 2206 East Mall, Vancouver, British Columbia, Canada, V6T 1Z3  c Department of Sociology, University of British Columbia, 6303 NW Marine Drive, Vancouver, British Columbia, Canada, V6T 1Z1  d B.C. Centre for Excellence in HIV/AIDS, 608 – 1081 Burrard Street, Vancouver, British Columbia, Canada, V6Z 1Y6  1 Present address: Department of Family Medicine, Queen's University, 220 Bagot Street, Kingston, Ontario, Canada, K7L 5E9  * Corresponding author:  Email: michael.otterstatter@bccdc.ca Phone: +1-604-707-2853 Fax: +1-604-707-2516 2  Abstract Background Although income assistance is an important source of support for low income individuals, there is evidence that adverse outcomes may increase when payments are disbursed on the same day for all recipients.  The objective of this study was to assess the temporal patterns and causal relation between population-level illicit drug overdose deaths and income assistance payments using daily mortality data for British Columbia over a period of five years. Methods Retrospective data on daily mortality due to illicit drug overdose, 2009 to 2013, were provided by the BC Coroners Service.  These data were analyzed using regression models and time series tests for causality in relation to dates of income assistance payments. Results 1343 deaths due to illicit drug overdose were reported in BC during 2009-2013; 394 occurred during cheque weeks (n=60) and 949 occurred during non-cheque weeks (n=202).  Average weekly mortality due to illicit drug overdose was 40% higher during weeks of income assistance payments compared to weeks without payments (P < 0.001).  Consistent increases in mortality appeared the day after cheque disbursement and were significantly higher for two days, and marginally higher after 3 days, even when controlling for other temporal trends.  Granger causality testing suggests the timing of cheque issue was causally linked to increased drug overdose mortality (P < 0.001).   Conclusions Our findings clarify the temporal relation and causal impact of income assistance payments on illicit drug deaths.  We estimate 77 avoidable deaths were attributable to the synchronized disbursement of income assistance cheques over the 3  five year period.  An important consideration is whether varying the timing of payments among recipients could reduce this excess mortality and the related demands on health and social services.   Keywords: Drug users; Harm reduction; Income assistance; Mortality; Overdose    4  Introduction   Income assistance is a form of social assistance that supports the basic needs of low income populations and prevents harms associated with poverty.1 However, there is a growing body of evidence suggesting negative individual and societal outcomes when income assistance is disbursed on the same day for all recipients. In British Columbia (BC), income assistance is provided as a scheduled cheque payment, generally on the last Wednesday of each month, by the Ministry of Social Development and Social Innovation.2 All persons in BC receiving income assistance are issued their cheque on the same day, usually at an interval of four weeks between cheques (8 months/year), but occasionally five weeks between cheques (4 months/year). Roughly 176,000 people receive income assistance in BC each year (72,000 persons expected to work or temporarily excused, 10,000 persons with persistent multiple barriers and 94,000 persons with disabilities).3    The harms of income assistance were documented in an early study on welfare cheque disbursement in Minnesota. Brunette et al. (2001) showed a correlation between cheque distribution and increased emergency service use, as well as jail, hospital, and alcohol receiving centre visits.4 Subsequent studies temporally linked social assistance with outcomes such as increased substance use and mortality after cheque issue.5–7 The monthly cycle of substance use attributed to the timing of social assistance became known as the “check effect”.8 Further studies on disability payments among veterans demonstrated that although overall levels of drug use may remain similar whether or not individuals receive social assistance, payments can alter the timing of drug use leading 5  to increased use in the period surrounding social assistance payments.8–10 In BC, previous studies show social assistance payments are associated with increased emergency department visits, emergency service responses, alcohol detoxification unit admissions, discharges from hospital against medical advice and likelihood of overdose and all-cause mortality.10–13  Previously reported increases in utilization of healthcare and social services following cheque day are largely correlative and lack the temporal resolution necessary to isolate the cheque effect. A prominent example by Phillips et al. (1999) considers national death data for the United States but only notes that those associated with substance abuse are more common during the first week of the month, coincident with the timing of federal payments.6 Hence, the precise timing and magnitude of this effect are unclear and causality remains an open question. We therefore undertook the current study to assess the daily temporal relationship and causal impact of income assistance payments on illicit drug overdose deaths across a five-year period using data from BC Coroners Service.  Methods Dates of illicit drug overdose death in BC for the five year period 2009 to 2013 were obtained from the BC Coroners Service. These data include all classifications of illicit drug overdose death (accidental, undetermined, suicide) and can include residents and non-residents of BC. Income assistance and disability assistance cheque issuance dates were obtained from the BC Ministry of Social Development and Social 6  Innovation.2 We generated from these data the primary outcome variable, daily overdose death counts, and identified days and weeks of cheque issue as well as basic calendar variables (weekday, month, year) for further analysis. Ethics approval was received from University of British Columbia Behavioural Research Ethics Board (H08-00333).    Statistical analyses We conducted several analyses to clarify broad- and fine-scale patterns in overdose mortality and to identify the impact of cheque issue independent of other temporal patterns. First, following previous work, we compared average weekly mortality due to overdose during cheque weeks vs. non-cheque weeks using a two-sample t-test, accounting for unequal variance.11  Second, we analyzed patterns of daily mortality using Poisson regression treating number of deaths as the dependent variable and weekday, month, year and cheque week (yes/no) as explanatory factors.  We explored all main effects and two-way interactions and removed interaction terms with P > 0.05.  The interaction of interest in this model was cheque week x weekday, indicating that the pattern of daily mortality differed between cheque vs. non-cheque weeks.  To identify significant differences in mortality we used linear contrasts to compare daily mortality between days (e.g., Thursdays following cheque issue vs. all other Thursdays) and combinations of days (e.g., weekends vs. weekdays).  We explored the possibility that mortality varied according to the number of weeks between cheques by running the above Poisson regression with a flag for ‘delayed’ cheque weeks, i.e., those with a five week interval between cheques instead of the usual four. Finally, we used time series 7  analysis on the residuals of our Poisson regression (i.e., weekday, month and year trends removed) to confirm the lagged effect of cheque issue on daily mortality and to test for causality. We used an autoregressive model including an exogenous variable for the timing of cheque issue (ARX model) and explored lagged effects up to seven days. By first removing temporal variation related to basic calendar variables (e.g., weekend effects), this analysis focused on fluctuations in mortality that occurred on, or shortly after, cheque issue. A Granger causality test was applied to the final model to assess whether or not cheque issuance was causally linked to overdose mortality.11 Granger causality does not identify a specific causal mechanism (this would require experimental manipulation), but provides a rigorous test of the directed influence of one variable on another.14 A significant test would indicate that cheque issuance is predictive of temporal patterns in overdose mortality, and not simply an association in time. In all regressions, model goodness-of-fit was assessed using standard tests (e.g., deviance/degrees-of-freedom for dispersion in Poisson models; Dickey-Fuller tests for stationarity in time-series models) and nested models were compared using Akaike information criterion (AIC). All analyses in this study used the complete (January 1 – December 31) years from 2009-2013.  Results A total of 1343 deaths due to illicit drug overdose were reported in BC during 2009-2013, with 394 occurring during cheque weeks (n=60) and 949 occurring during non-cheque weeks (n=202). The average number of overdose deaths was 40% higher during cheque weeks compared to non-cheque weeks (mean ± SD: 6.56 ± 3.06 deaths 8  vs. 4.70 ± 2.24 deaths; P < 0.001). Cheque weeks occurring after an interval of five weeks since last cheque tended to show even more overdose related deaths (7.36 ± 3.05 deaths, n = 22) than cheque weeks occurring after an interval of four weeks (6.19 ± 3.01 deaths, n = 37), although the difference between overdoses after five-week and four-week social assistance cycles was not significant (P = 0.15).    Figure 1 shows patterns of daily mortality during cheque and non-cheque weeks.  There was no difference in the number of deaths on cheque Wednesdays compared to all other Wednesdays (0.69 vs. 0.61 daily deaths, P = 0.50). However, there was a significantly greater number of deaths on Thursdays one day after cheque issue compared to all other Thursdays (1.24 vs. 0.55 daily deaths, P < 0.001) and on Fridays two days following cheque issue compared to all other Fridays (1.01 vs. 0.71 daily deaths, P = 0.007). This differing pattern of daily mortality between cheque weeks and non-cheque weeks was statistically significant in our regression model, as shown by the cheque week x day interaction in Table 1.  Combing average daily deaths on Thursdays and Fridays following cheque issue shows that mortality increased to nearly double that of all other days in our dataset (1.26 vs. 0.68 daily deaths, P < 0.001).  The magnitude of this ‘cheque effect’ appeared more pronounced when the interval between cheques was five weeks instead of four weeks (1.44 vs. 1.16 daily deaths), although again the difference between four- and five-week cycles was not significant (P = 0.23).    Compared to non-cheque weeks, average daily deaths tended to remain higher for the entire seven days of a cheque week.  Although our day-to-day comparisons showed 9  significant differences for 2 days beyond cheque issue, further time-series analysis found a significant difference (P = 0.045) at 3 days post cheque issue as well.  Across 2009-2013, a total of 200 overdose deaths were observed during the three days following cheque issue, compared to 123 deaths expected based on the rate during the corresponding Thursday-Saturday of non-cheque weeks. This suggests roughly 77 extra deaths were attributable to the disbursement of income assistance cheques over five years.  The Granger causality test in our time series analysis was highly significant (P < 0.001), suggesting that the timing of cheque issue was indeed causally linked to drug overdose mortality.    Figure 2 shows the pattern of daily overdose mortality across an entire month. Increased mortality is apparent in the days following cheque issue and was consistent regardless of month (non-significant cheque week x month effect, P = 0.18) or year (non-significant cheque week x year effect, P = 0.18).  During non-cheque weeks, average daily deaths were greater on weekends compared to the rest of the week (G = 19.0, P < 0.001).  This effect was not visible during cheque weeks due to the already elevated death rate resulting from the Thursday-Friday cheque effect.  It is noteworthy that cheque weeks occurring after a five week interval typically showed fewer deaths in the days leading up to cheque issue, but then showed an increased number of deaths on cheque Wednesday, in contrast to cheque weeks occurring after a four week interval.    Discussion 10  In our study, we found consistent increases in overdose mortality during the days immediately following issuance of income assistance cheques.  Across a five year period in British Columbia, roughly twice the number of overdose deaths occurred on the two days following cheque issuance compared to the corresponding two days during non-cheque weeks.  A marginally significant increase in mortality was also observed at three days post cheque issuance. Our analysis suggests that roughly 77 deaths were attributable to the disbursement of income cheques over the five year study period. Given that only four percent of the BC population receives temporary employment and disability assistance, these deaths represent a notable burden of potentially preventable mortality.3 A novelty of this study was the statistical testing for Granger causality, which reinforces the conclusion that observed increases in overdose mortality are linked to the timing of income assistance cheques, which are disbursed each month to all recipients on a single day.  Our finding of lagged overdose mortality following cheque issuance adds new and important detail to existing studies.  Previous work has shown that non-fatal overdoses at InSite, the local supervised injection facility, doubled across a three-day period beginning with the day of cheque issue, after adjusting for the number of injections at the site.15  Similarly, both mental health apprehensions by police and mental health and substance–related emergency department presentations show significant increases across the seven-day period starting on the day of cheque disbursement.16  Such studies, analysing aggregate days following cheque issuance, do not permit assessment of daily patterns such as lagged increases in overdose.  Our results show 11  that analysis of daily patterns is essential for assessing the true magnitude and extent of the effect, particularly since the timing of fatal overdoses may lag behind cheque issuance and occur against a background of unrelated trends (e.g., weekend effects).  One previous study using daily data found that admissions to a sobering unit for publicly intoxicated individuals were elevated on each of the three consecutive days starting with cheque issue and, like our study, found that these peaked on the day following cheque issue.17  We add to this body of work, which is largely focused on non-fatal events, by showing that a critical outcome documented through highly specific data, i.e., coroner diagnosed illicit drug overdose deaths, also increases in a consistent and significant manner following cheque issue.  Rosen (2011) concluded that “…people vulnerable to substance abuse are more likely to use drugs and/or alcohol around the time they receive funds, especially people whose expenditures are limited by poverty”.8  Moreover, increased high-risk behaviour, such as binge use, poly-substance use and higher doses may contribute to adverse outcomes during this time.15  Interestingly, we observed no increase in mortality on the actual day of cheque disbursement in the majority of cheque weeks; yet, in ‘delayed’ cheque weeks, with cheque issuances separated by five weeks instead of the usual four weeks, mortality appeared elevated on the day of cheque disbursement as well.  This may suggest not only an increase in drug use and high-risk behaviour in the time following cheque issuance, but also a further reduction of tolerance of opioids over five weeks compared to four weeks. Reduced individual tolerance was previously hypothesized as a contributing factor to increased non-fatal overdoses at Vancouver’s 12  supervised injection site.15    Strengths Our analysis of five years of comprehensive coroners data provides a large sample (N=1343) of illicit drug overdose deaths for assessing temporal changes in mortality. Time series analysis and causality testing add validity to the conclusion that cheque issuance has a causal impact on drug overdose mortality, rather than being purely associational. Previous studies have focused on subpopulations, for example veterans, or people using medical or addiction services such emergency departments or supervised injection sites. Our study adds additional breadth to this literature by being a fully population-based study.  We also observed certain patterns in mortality occurring independently of cheque issuance, e.g., increased daily deaths on weekends compared to the rest of the week. Previous reports on drug poisoning deaths suggest such weekend effects may result from increased recreational drug use and weekend events such as raves, parties and nightclubs.18  Limitations Although our results are consistent with those from previous studies, one limitation is necessarily the observational nature of this work.  Our results do not imply, for example, that all overdose deaths following cheque issuance were among individuals receiving income assistance.  Similarly, our analysis of population-level data cannot clarify the mechanism by which income assistance increases an individual’s risk of overdose, or the exact time at which an overdose occurred.  Nevertheless, other studies on this topic 13  have identified plausible pathways linking the influx of income to increased expenditure on drug use to patterns of high risk drug use.  Furthermore, it may be that the timing of cheque issue serves as a cue for increased use of illicit drugs at a community level, for example through social events that increase risk of overdose, and that the impact of cheque issue extends to those not receiving income assistance.19  We were unable to investigate differences according to the specific substance causing death, or the contributions of individual-level factors possibly related to overdose, such as age, gender, duration of drug use and contribution of other substances (e.g., alcohol).  Given that patterns of substance use in BC vary by age, gender, and geography, future research could explore the relation between these covariates and temporal trends in illicit drug overdose deaths preceding and following disbursement of income assistance cheques.20  Conclusions Our study adds support and additional detail to a growing body of evidence showing increased deaths in the days following income assistance payments.  There is, therefore, a strong rationale for examining alternative disbursement schedules experimentally (e.g., varying the timing of cheque issuance among recipients, issuing smaller and more frequent cheques, etc.) in order to reduce this avoidable mortality and the demands on health and social services.  Given the consistent increase in overdose deaths following cheque issue, there is a clear need for harm reduction and support 14  services to be available during this time to mitigate adverse outcomes related to substance use and income assistance payments.    Acknowledgements We gratefully acknowledge the BC Coroners Service for providing date of death data on illicit drug overdoses and for helpful comments on the manuscript. 15  Tables  Table 1. Regression model statistics describing the effect of cheque week on average daily deaths due to illicit drug overdose. Variable DF LR χ2   P-value Month 11 12.77    0.309 Day 6 20.7    0.002 Year 4 38.76 < 0.001 Cheque week 1 25.95 < 0.001 Cheque week x Day 6 14.03    0.029 * non-significant interactions not shown   16  Figures    Figure 1. Average daily mortality due to illicit drug overdose during weeks of cheque issue versus weeks without cheque issue.  Cheque issue occurred on Wednesdays.  Values are adjusted for the effects of month and year based on our regression model (see Methods). Asterisks indicate significant differences between cheque and non-cheque weeks.17    Figure 2. Average daily deaths due to illicit drug overdose during a four week period, including the timing of issuance for income assistance cheques.  The two week period prior to cheque issuance is shown for both 4 week cheque cycles (open bars) and 5 week cheque cycles (light grey bars). The solid trend line represents a two day moving average of daily deaths. 18  References 1.  Walker R. Social Security and Welfare: Concepts and Comparisons. Berkshire: Open University Press; 2005. 2.  Income Assistance and Disability Assistance Cheque Issue Dates. BC Ministry of Social Development and Social Innovation. http://www.eia.gov.bc.ca/publicat/bcea/Cheqissu.htm. Published 2015. Accessed July 23, 2015. 3.  BC Employment and Assistance Summary Report - May 2015. Victoria, BC; 2015. http://www.sd.gov.bc.ca/research/15/06-may2015.pdf. Accessed July 24, 2015. 4.  Brunette DD, Kominsky J, Ruiz E. Correlation of emergency health care use, 911 volume, and jail activity with welfare check distribution. Ann Emerg Med. 1991;20(7):739-742. 5.  Shaner A, Eckman TA, Roberts LJ, et al. Disability income, cocaine use, and repeated hospitalization among schizophrenic cocaine abusers--a government-sponsored revolving door? N Engl J Med. 1995;333(12):777-783. 6.  Phillips DP, Christenfeld N, Ryan NM. An increase in the number of deaths in the United States in the first week of the month--an association with substance abuse and other causes of death. N Engl J Med. 1999;341(2):93-98. 7.  Halpern SD, Mechem CC. Declining rate of substance abuse throughout the month. Am J Med. 2001;110(5):347-351. 8.  Rosen MI. The “check effect” reconsidered. Addiction. 2011;106(6):1071-1077. 9.  Rosenheck R, Frisman L. Do public support payments encourage substance abuse? Health Aff. 1996;15(3):192-200. 10.  Swartz JA, Hsieh C, Baumohl J. Disability payments, drug use and representative payees: an analysis of the relationships. Addiction. 2003;98(7):965-975. 19  11.  Verheul G, Singer SM, Christenson JM. Mortality and morbidity associated with the distribution of monthly welfare payments. Acad Emerg Med. 1997;4(2):118-123. 12.  Anis AH, Sun H, Guh DP, Palepu A, Schechter MT, O’Shaughnessy M V. Leaving hospital against medical advice among HIV-positive patients. CMAJ. 2002;167(6):633-637. 13.  Riddell R, Riddell C. Welfare Checks, Drug Consumption, and Health: Evidence from Vancouver Injection Drug Users. J Hum Resour. 2006;41(1):138-161. 14.  Barrett AB, Barnett L. Granger causality is designed to measure effect, not mechanism. Front Neuroinform. 2013;7:6. http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3635029&tool=pmcentrez&rendertype=abstract. Accessed September 15, 2015. 15.  Zlotorzynska M, Milloy M-JS, Richardson L, et al. Timing of income assistance payment and overdose patterns at a Canadian supervised injection facility. Int J Drug Policy. 2014;25(4):736-739. 16.  Pickett TA, Stenstrom RJ, Abu-Laban RB. Association between mental health apprehensions by police and monthly income assistance (welfare) payments. Can J Psychiatry. 2015;60(3):146-150. 17.  Li X, Sun H, Marsh DC, Anis AH. Impact of welfare cheque issue days on a service for those intoxicated in public. Harm Reduct J. 2007;4(1):12. 18.  Pennsylvania State Coroners Association. Report on Overdose Death Statistics: 2014.; 2015. http://www.pacoroners.org/Uploads/Pennsylvania_State_Coroners_Association_Drug_Report_2014.pdf. Accessed July 23, 2015. 19.  Small W, Shoveller J, Moore D, Tyndall M, Wood E, Kerr T. Injection drug users’ access to a supervised injection facility in Vancouver, Canada: The influence of operating policies and local drug culture. Qual Heal Res. 2011;21(6):743-756. 20  20.  Kuo M, Shamsian A, Tzemis D, JA B. A drug use survey among clients of harm reduction sites across British Columbia, Canada. Harm Reduct J. 2014;11 (1):13.  


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items