Estimates of compliance with Canada’s guidelines for low and moderate risk alcohol consumption: the importance of adjustment for underreporting in self-report surveys

Abstract

Alcohol is a leading cause of premature mortality; however, awareness of this and of some specific risks is low. Survey-based estimates of drinking at risk levels suffer from substantial underreporting. We show that alcohol use reported in the 2019 Canadian Alcohol and Drug Survey (CADS) accounted for just 38.06% of recorded alcohol consumption. This contributes to alcohol’s risks being minimized, by researchers, the public, and policymakers. The new Canada’s Guidance on Alcohol and Health (CGAH) defines “moderate risk” as 3 to 6 drinks/week for men and women. Employing published methods to correct for underreporting in the CADS, we estimate that in 2019 the proportion of drinkers at moderate risk of long-term harm to be 50.43% (up from 23.34% without adjustment). We further estimate that, collectively, these drinkers consumed 90.17% of all drinks consumed that year. Similarly, 92.82% of drinks were consumed on days when upper limits for short-term harm (2 drinks/day) were exceeded (up from 65.02% without adjustment). We conclude that adjustments for underreporting of alcohol use in Canada need to be incorporated routinely in public health monitoring. This might help mitigate the widespread underestimation of risky alcohol use as well as the neglect of this public health issue by policymakers.

Supplementary Information

The online version contains supplementary material available at 10.17269/s41997-023-00781-6.

Alcohol use in Canada was estimated to be responsible for over 17,000 deaths and 117,000 hospital admissions at a cost of $19.7 billion in 2020 (Canadian Substance Use Costs & Harms Scientific Working Group, 2023). One strategy to reduce health risks from drinking is to encourage consumers to restrict their consumption within “low risk” limits. Canada’s 2011 low-risk drinking guidelines (LRDGs) were generous by international standards, recommending women drink up to 10 standard drinks (SDs) per week (1 SD = 17.05 mL ethanol) and men up to 15 per week (Butt et al., 2011). The proportion of Canadians exceeding these limits has been widely reported by public health authorities (Canadian Institute for Health Information, 2017). However, despite using large representative population samples, estimates of LRDG compliance from national self-report surveys suffer from massive underreporting, typically covering only 30 to 40% of known age 15+ per capita alcohol consumption from recorded alcohol sales (Stockwell et al., 2018a, 2018b; Zhao et al., 2015). We revisit this problem in light of the new, stricter guidance on alcohol and health for Canadians (Paradis et al., 2023).

Methods are available to adjust for underreporting

Given the scale and seriousness of alcohol-related harm in Canada, we submit that adjustment for underreporting of consumption should be applied routinely for public health monitoring purposes, with age 15+ per capita alcohol consumption (PCAC) estimates from sales data as the gold standard. Such adjustments are used by the World Health Organization in its Global Burden of Disease reports (Stockwell et al., 2018a, 2018b). We have previously described the necessary methods (Stockwell et al., 2014) and applied them in the Canadian context (Zhao et al., 2015). After adjustment, we previously estimated that between 2008 and 2010, 27.3% of Canadians exceeded the generous 2011 guidelines for reducing risk of long-term harms (e.g. liver cirrhosis) and 38.6% for risk of short-term harms (e.g. poisonings), compared with unadjusted estimates of 6.8% and 16.7% respectively (Zhao et al., 2015). We provide new estimates here of risky alcohol consumption using data from the 2019 Canadian Alcohol and Drug Survey (CADS) (Statistics Canada, 2020) and reference these against both the recent new Canada’s Guidance on Alcohol and Health (CGAH) (Paradis et al., 2023) and previous 2011 LRDG (Butt et al., 2011).

How we applied correction factors for underreporting in the 2019 CADS

Estimates of alcohol consumption by age and sex were taken from the 2019 CADS (Statistics Canada, 2020), adjusted for patterns of underreporting and then “uplifted” to match known total age 15+ PCAC estimated from sales data (Statistics Canada, 2021). Age- and sex-specific correction factors for underreporting previously calculated from the 2008–2010 CADUMS (Canadian Alcohol and Drug Use Monitoring Survey) for drinking quantities and frequencies were first applied before uplifting to match sales data (Stockwell et al., 2014; Zhao et al., 2015). These factors were based on analyses of responses to a question about whether respondents drank alcohol the day immediately before the interview (“Yesterday”) and the reported amounts consumed. The underlying assumptions were (a) recall of drinking “yesterday” is accurate and (b) observed probabilities of alcohol consumption “yesterday” provide accurate estimates of mean drinking frequencies for sub-groups of drinkers. Since these correction factors reflect patterns of recall and understanding of language used in survey questions, it was assumed these proportional adjustment factors would still apply in 2019 as the CADS has no Yesterday question. The resulting adjustments are similar for men and women but differ by age (Stockwell et al., 2014; Zhao et al., 2015).

The 2019 CADS was a national telephone survey of 10,293 persons aged 15 years or older living in Canada’s 10 provinces (Statistics Canada, 2020).

Measures of alcohol consumption

Self-reported (CADS)

The quantity and frequency (QF) approach was used to estimate total alcohol consumption in 2019 of each drinker, i.e. self-reported drinking frequency multiplied by typical number of standard drinks per drinking day in past year (see Supplementary Materials, Appendix A). Mean consumption (SDs/person/year) was calculated by age (15–19, 20–24, 25–44, 45–64, and 65+) and sex (self-identified males and females). Unlike the 2008–2010 CADUMS, responses to the typical quantity of alcohol consumed on a drinking day were capped at 11 drinks per day by Statistics Canada compared with 50 drinks per day in the CADUMS.

Recorded sales

Statistics Canada (2020) data (Statistics Canada, 2020) were accessed on litres of pure alcohol per person aged 15+ across the 10 provinces in the fiscal year 2018/2019 in total and also by each of four main beverage types: beer, coolers, spirits, and wine.

Adjustment for underreporting of alcohol consumption

Non-compliance with (a) the new CGAH (Paradis et al., 2023) and (b) the 2011 LRDGs was estimated with and without adjustment for underreporting. This was done for “risky drinkers” (those not complying with guidelines) and “risky drinks” (consumed on non-compliant days).

Risky drinking for long-term health harms is defined as (a) 7+ drinks/week for men and women in the 2023 CGAH and (b) 16+ drinks/week for men and 11+ drinks/week for women in the 2011 LRDGs. “Risky” drinks for long-term health harms were counted as all those consumed by these “risky drinkers”. Risky drinking for short-term health harms is defined as (a) 3+ drinks on 1 day in the 2023 CGAH and (b) 4+ /5+ drinks on 1 day for women/men in the 2011 LRDG. All drinks consumed on such days were counted as “risky drinks” for short-term harm.

Statistical analysis

Estimates were based on the weighted sample with expansion weights recalculated and rescaled to the sample size (Statistics Canada, 2020). Z-tests were used to examine the differences in estimates by age and sex groups (Pagano & Gauvreau, 2000) applying two-sided P-value < 0.05 significance criteria. SAS SURVEYMEANS, SURVEYFREQ, and SURVEYREG procedures were used to account for survey sample design effects (SAS Institute, 2016).

Our adjusted estimates of compliance with Canada’s new and old alcohol guidelines

The 2019 CADS sample was biased towards older people and contained more females than males (Table 1). Age 15+ per capita consumption calculated from the unadjusted 2019 CADS was only 3.06 L of ethanol in 2019 compared to 8.04 L estimated by Statistics Canada for 2018/2019, i.e. just 38.06% of recorded alcohol sales.

Table 1.

Percentage (%) estimates and 95% confidence intervals (CI) of lifetime abstainers, former drinkers, and current drinkers by age and sex in 2019

Variable	N-sample †	Lifetime abstainers ‡		Former drinker ‡		Current drinker ‡		X2, DF, P-value *
		%	95% CI	%	95% CI	%	95% CI
Age in years
15–19	272	51.45	34.72–68.17	2.29	0.00–6.10	46.25	30.81–61.70	397.05, 8, < 0.001
20–24	319	10.25	7.66–12.85	5.36	1.76–8.96	84.38	78.67–90.08
25–44	2815	10.78	7.29–14.27	8.08	7.46–8.71	81.12	77.86–84.38
45–64	3771	9.74	7.64–11.84	10.86	9.24–12.48	79.38	76.35–82.40
65 +	3116	9.95	8.40–11.51	18.24	15.61–20.88	71.79	69.48–74.11
Sex
Male	4669	12.18	8.36–15.99	9.54	8.79–10.28	78.27	74.80–81.74	17.55, 2, < 0.001
Female	5624	13.88	11.08–16.68	11.44	9.92–12.96	74.67	70.95–78.40
Canada	10,293	13.03	9.79–16.28	10.50	9.49–11.51	76.45	72.92–79.98

Note: †, unweighted sample. ‡, weighted estimates. Lifetime abstainers were defined as those who reported never consuming one drink in their lives; former drinkers were those who reported drinking before but not in last year; and current drinkers were those who reported at least one drink in past year. *Rao-Scott chi-square tests statistic and P-value

It was estimated that 50.43% and 30.70% of Canadian drinkers exceeded the 2023 and 2011 drinking guidelines respectively for long-term harm after adjustment for underreporting (Table 2). Many more men than women (62% versus 38%) regularly exceeded the new limits for weekly consumption. Breakdowns by age are provided in Appendix B (see Supplementary Materials). The amount of alcohol consumed by people exceeding these weekly limits was estimated to comprise 90.17% (2023 CGAH) and 74.40% (2011 LRDG) of all alcohol consumed in 2019 (Table 2).

Table 2.

Weighted estimates of %’s of risky drinkers and risky drinks for long-term harm from the 2019 CADS as defined by Canada’s old 2011 and new 2023 alcohol and health guidelines, with and without adjustment for underreporting

Variable	N †	Unadjusted risky drinkers			Adjusted risky drinkers ‼			Unadjusted risky drinks				Adjusted risky drinks ‼
		% ‡	95% CI	z-test P	% ‡	95% CI	z-test P	N ††	% ‡‡	95% CI	z-test P	N ††	% ‡‡	95% CI	z-test P
Risk defined by Canada’s 2023 Alcohol and Health Guidelines (7+ drinks/week)
Sex
Male	3693	28.70	26.53–30.88	< 0.001	62.20	58.58–65.83	< 0.001	21,174	77.75	77.21–78.29	< 0.001	57,448	94.29	94.10–94.47	< 0.001
Female	4252	17.88	15.30–20.45	Ref	38.42	32.81–44.02	Ref	13,923	66.05	65.24–66.86	Ref	33,732	82.28	81.87–82.70	Ref
Canada	7945	23.34	21.50–25.19		50.43	45.78–55.09		35,096	73.44	72.98–73.90		91,180	90.17	89.98–90.36
Risk defined by Canada’s 2011 Low Risk Drinking Guidelines (11+ /16+ drinks/week for women/men)
Sex
Male	3693	8.17	6.43–9.91	= 0.459	32.81	28.85–36.78	< 0.001	21,174	42.15	41.51–42.80	< 0.001	57,448	74.93	74.58–75.27	< 0.001
Female	4252	8.63	6.55–10.71	Ref	28.56	25.80–31.31	Ref	13,923	44.55	43.70–45.40	Ref	33,732	73.39	72.85–73.82	Ref
Canada	7945	8.40	7.13–8.68		30.70	28.00–33.41		35,096	43.03	42.52–43.55		91,180	74.40	74.12–74.68

Note: †, drinkers in past year unweighted. ‼, adjusted first for estimated degree of underreporting by age and sex, then further adjusted to the level of official sales. ‡, % of weighted sample. ††, total drinks per week unweighted. ‡‡, % of weighted sample

It was estimated that 49.90% of all drinkers exceeded new guidelines for reducing short-term harm (3+/day) at some point during the previous year (Table 3). Men were just as likely as women to exceed these differential limits. The proportion of drinks consumed on days when 2023 limits for short-term harm were exceeded comprised 92.82% of all drinks consumed in 2019 (Table 3).

Table 3.

Weighted estimates of %’s of risky drinkers and risky drinks for short-term harm from the 2019 CADS as defined by Canada’s old 2011 and new 2023 alcohol and health guidelines, with and without adjustment for underreporting

Variable	Total N †	Unadjusted risky drinkers			Adjusted risky drinkers ‼			Unadjusted risky drinks				Adjusted risky drinks ‼
		% ‡	95% CI	z-test P	% ‡	95% CI	z-test P	Total N ††	% ‡‡	95% CI	z-test P	Total N ††	% ‡‡	95% CI	z-test P
Risk defined by Canada’s 2023 Alcohol and Health Guidelines (3+ drinks on 1 day)
Sex
Male	3693	46.13	44.17–48.09	< 0.001	80.47	70.34–90.60	= 0.992	9753	71.12	70.26–71.97	< 0.001	19,094	93.38	93.05–93.72	< 0.001
Female	4252	33.24	30.00–36.48	Ref	80.48	77.95–83.02	Ref	8880	57.21	56.15–58.27	Ref	16,425	92.05	91.63–92.48	Ref
Canada	7945	39.75	37.75–41.76		80.47	74.58–86.37		18,633	65.02	64.35–65.70		35,520	92.82	92.55–93.08
Risk defined by Canada’s 2011 Low Risk Drinking Guidelines (4+/5+ drinks on 1 day for women/men)
Sex
Male	3693	12.93	10.27–15.58	< 0.001	49.44	40.90–57.99	= 0.412	9753	35.03	34.13–35.93	< 0.001	19,094	74.99	74.40–75.58	< 0.001
Female	4252	17.93	15.90–19.56	Ref	50.36	45.41–55.31	Ref	8880	39.25	38.20–40.29	Ref	16,425	73.44	72.74–74.13	Ref
Canada	7945	15.30	13.17–17.43		49.90	43.19–56.61		18,633	36.88	36.19–37.56		35,520	74.33	73.88–74.78

Note: †, drinkers in past year unweighted. ‼, adjusted first for estimated degree of underreporting by age and sex, then further adjusted to the level of official sales. ††, total drinks per drinking day unweighted. ‡‡, % of weighted sample

Implications of our findings for public health monitoring in Canada

It is necessary to adjust estimates of risky alcohol use from self-report surveys upwards to reflect actual population consumption as recorded from sales data to compensate for substantial underreporting in self-report surveys. It should be noted that the estimates here are conservative as they do not account for unrecorded consumption which adds 10% to per capita consumption in Canada (WHO, 2018) and, further, the 2019 CADS capped self-reported daily consumption at 11 drinks/day, well below consumption of people with severe alcohol use disorders (Stockwell et al., 2018a, 2018b).

Despite the conservative nature of our estimates, the adjustments for underreporting implemented suggest that the great majority of alcohol consumed in Canada is consumed in a way which puts the drinker at risk of both short- and long-term harm. We estimated that over half of Canadian drinkers exceed “moderate risk” levels for long-term harm and over 80% do so for short-term harm. Further, as many as 90% of all the drinks consumed in Canada in the year 2019 were consumed by moderate risk drinkers, while 93% of all the drinks consumed that year were drunk on days when new guideline levels for short-term harm were exceeded.

These findings have important implications for alcohol policy in Canada. Canada’s new guidelines advise people who drink that “less is best”, that there is a continuum of risk with “low” and “moderate” risk levels being identified along that continuum (Paradis et al., 2023). Even at the “moderate risk” level (a 1% increase in mortality risk from alcohol-related causes), a substantial majority (> 90%) of all drinks consumed in Canada in 2019 were estimated to have been drunk by people exceeding limits, whether for short- or long-term harm. This strengthens the case for stronger policies to reduce alcohol-related harms and undermines the popular idea that only a small number of people put their health at risk through their alcohol consumption. We further recommend that to capture the full range of Canadian patterns of alcohol consumption, future national surveys reintroduce the Yesterday Method questions to better account for patterns of underreporting and, further, a more realistic cap of at least 20 drinks per day is used in the data reported. We note that among people with severe alcohol use disorders attending Managed Alcohol Programs in Canada, even average daily consumption was approximately 21 standard drinks (Stockwell et al., 2021).

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

TS and JZ designed the analysis, JZ conducted the analysis, TS wrote the first draft of the paper, JZ edited it, and both authors approved the final draft.

Funding

We acknowledge funding from Health Canada’s Office of Drug Policy and Science for the completion of the work reported in this article.

Data availability

Available in public domain from Statistics Canada.

Code availability

SAS software v9.4.

Declarations

Ethics approval

Not applicable.

Consent to participate

Not applicable.

Consent for publication

Not applicable.

Conflict of interest

TS declares receipt of research funding, travel expenses, and personal fees for work commissioned by Alko, Finland and Systembolaget, Sweden, for the conduct of research to evaluate the public health impacts of alternative alcohol policies in those countries. These are state-owned government alcohol monopolies that report to their respective Ministries for Health and Social Affairs and have a mandate to mitigate harms from alcohol consumption. JZ declares no conflicts of interest.