Although the proportion of adults aged 55 and older is larger than ever among the Canadian population, older adults have drawn little attention in alcohol research despite their unique relationship with alcohol. Normal physiological changes caused by aging reduce alcohol tolerance, which usually translates into a decrease in alcohol consumption with increasing age (Graham, Reference Graham1986; Menninger, Reference Menninger2002; Zimberg, Reference Zimberg1978). Older adults often experience more harmful effects than younger adults when consuming the same quantities of alcohol, due to higher and longer-lasting blood alcohol concentrations; in addition, they are more likely to suffer from chronic diseases and use medication that may interact with alcohol (Dufour & Fuller, Reference Dufour and Fuller1995; Tanaka, Reference Tanaka2003; Whelan, Reference Whelan2003). Accordingly, Canadian cross-sectional data show that the quantity of alcohol consumed per occasion tends to decrease with age, whereas the frequency of drinking tends to increase (Demers & Poulin, Reference Demers, Poulin, Adlaf, Begin and Sawka2005; Graham, Carver, & Brett, Reference Graham, Carver and Brett1995).
While older adults are known to drink differently than do younger adults, little is known about changes in their alcohol consumption over time. In fact, with regard to drinking behaviours, aging does not appear to be as important a factor as attitude, generation, and time period (Glynn, Bouchard, LoCastro, & Laird, Reference Glynn, Bouchard, LoCastro and Laird1985; Levenson, Aldwin, & Spiro, Reference Levenson, Aldwin and Spiro1997; Moore et al., Reference Moore, Gould, Reuben, Greendale, Carter and Zhou2005). However, social contexts and attitudes towards alcohol consumption have changed drastically in recent decades, paralleling increasing levels of education and standards of living and the massive entry of women into the labour force. In addition, adults aged 55 and older in the 2000s have displayed different characteristics from previous cohorts of the same age. They represent the first generation of men and women to have consumed alcohol throughout their lifetime in a context of gender equality and to report a higher level of alcohol consumption during their youth and adult life (Racine & Sawka, Reference Racine, Sawka, Adlaf, Begin and Sawka2005). Alcohol consumption per capita in Canada increased up to the years 1975–1980 and then declined until the 1990s, with a subsequent increase since 1996 (Ramstedt, Reference Ramstedt2004). Canadian demographic projections predict a stronger reversal of the age pyramid in the future, and epidemiologists foresee an increase in the absolute number of drinkers among older adults (Cornman & Kingson, Reference Cornman and Kingson1996; Patterson & Jeste, Reference Patterson and Jeste1999; Statistics Canada, 2010). Not only are we currently experiencing an increase in the number of older adults, but the latter also differ socio-demographically from those of previous cohorts of the same age, and they live in a social and environmental context that is more conducive to alcohol consumption in general.
Accordingly, we devised this study with the aim of determining whether the contextual (time period) changes unique to the decade spanning the 1990s and the beginning of the 2000s had a significant effect on rates of drinkers and drinking profiles among older Canadians when controlling for the socio-demographic changes that also occurred during that period.
Contextual Changes Associated with Alcohol
Contextual changes refer to changes in the physical environment as well as in values and social norms that evolve in a person’s interaction with the environment (Cohen, Scribner, & Farley, Reference Cohen, Scribner and Farley2000). Public health interventions are based primarily on the ecological theory stating that contextual factors are critical determinants of individual behaviour (Cohen et al., Reference Cohen, Scribner and Farley2000). Three contextual changes have occurred since the 1990s that may have promoted a marked increase in drinking among the general population: (1) the marketing of alcoholic products, (2) the availability of alcohol, and (3) the media coverage of the protective effects of moderate drinking on cardiovascular health. This last element concerns older adults in particular because they are more likely to experience the protective cardiovascular effects of moderate drinking (Dixon, Dixon, & O’Brien, Reference Dixon, Dixon and O’Brien2003; Ellison, Reference Ellison2002). Alcohol consumption per capita and sales of alcoholic beverages have increased in Canada since the 1990s (Statistics Canada, 1997d, 2005); alcohol consumption among older adults is expected to have followed this trend and similarly increased.
Alcohol Marketing. Alcohol marketing is a growing phenomenon that is expected to have encouraged an increase in overall alcohol consumption in the Canadian population since the 1990s. The most common promotional techniques used for alcoholic beverages include logos on packaging, event sponsorship, newsletters providing information on alcohol and food pairings, air-mile rewards for alcohol purchases, and radio and television advertising (Giesbrecht, Ialomiteanu, Anglin, & Adlaf, Reference Giesbrecht, Ialomiteanu, Anglin and Adlaf2007). Electronic media are now used as much as print media for product promotion.
In addition, alcohol advertising is increasingly specialized and tailored to specific products and clienteles (Giesbrecht, Reference Giesbrecht2000). For example, spirits producers tend to stress alcohol equivalence, that is, the similar alcohol content of a standard glass of spirits, beer, or wine. Wine producers, however, focus on drinking with meals, socializing, and the beneficial effects of moderate alcohol consumption, including the promotion of the Mediterranean diet. Some beer producers promote beverages that are low in calories (although not necessarily in alcohol) to encourage consumption by women and people who are trying to control their weight. Beer marketing generally associates the product with sports and recreational activities, relaxing after a day of work, male bonding, youth, vitality, and energy (Giesbrecht, Reference Giesbrecht2000).
As a marketing technique, alcohol advertising is known to be associated with an increase in overall alcohol consumption and binge drinking in exposed populations (Saffer, Reference Saffer2002). We do not know whether alcohol-related advertising increased over the 1990s, creating greater exposure of the population to alcohol publicity; however, it is clear that alcohol marketing is evolving and becoming increasingly effective, thus likely encouraging drinking among the Canadian population.
Alcohol Availability. The availability of alcoholic products in Canada has increased significantly over the past 20 years, which may also have translated into increased overall alcohol consumption in the population. According to the availability theory (Bruun et al., Reference Bruun, Edwards, Lumio, Makela, Pan and Popham1975; Single, Reference Single, Chaudron and Wilkinson1988; Stockwell & Gruenewald, Reference Stockwell, Gruenewald, Heather and Stockwell2004): (1) the greater the availability of alcoholic beverages in a society, the higher the average alcohol consumption in its population will be, (2) the higher the average alcohol consumption, the higher the number of binge drinkers, and (3) the higher the number of binge drinkers, the greater the extent of alcohol-related social and health problems. In Canada, the evolution of alcohol availability since the 1990s has consisted of relatively stable economic availability (Single, Truong, Adlaf, & Ialomiteanu, Reference Single, Truong, Adlaf and Ialomiteanu1999) and significantly heightened physical availability (Giesbrecht, Reference Giesbrecht2000).
Economic availability refers to alcoholic beverage prices and taxes. Since the 1990s, economic availability has remained relatively constant, with prices increasing gradually alongside general market prices. The actual price of alcohol decreased from the years following the Second World War until the mid-1980s. Since then, the price of alcohol has remained at par with the prices of other consumer goods or slightly higher (Single et al., Reference Single, Truong, Adlaf and Ialomiteanu1999). Canadian alcohol producers tend to adopt a protectionist stance and support the idea of a minimum price set by law (Giesbrecht, Reference Giesbrecht2000). Higher prices (and taxes) would be expected to result in less drinking and a reduction in alcohol-related harm. However, studies on the topic show variations in the relationship between price and alcohol consumption, which is not as clear and unambiguous as might be supposed in theory (Centre for Addiction and Mental Health, 2004; Giesbrecht, Reference Giesbrecht2000). It is therefore difficult to speculate on the probable impact of the evolution of alcohol’s economic availability since the 1990s on the drinking habits of the Canadian population.
Physical availability is primarily defined by the number and types of outlets, the number and range of alcoholic products sold, the days and hours of availability, and the people to whom alcohol may be sold (e.g., the minimum age required to purchase alcoholic beverages, Stockwell & Gruenewald, Reference Stockwell, Gruenewald, Heather and Stockwell2004). Physical availability has clearly increased in Canada since the 1990s, and this is mainly due to increased outlet density (Giesbrecht, Reference Giesbrecht2000). For example, the number of provincial and territorial liquor authority stores and agencies in Canada (excluding Alberta) was 1,504 stores and 862 agencies in 1995–1996 and 1,546 stores and 1,949 agencies in 2004–2005 (Statistics Canada, 1997d, 2005). During this nearly 10-year interval, the number of alcohol outlets increased throughout Canada in four provinces in particular: Newfoundland/Labrador (from 96 stores and agencies in 1995–1996 to 139 in 2004–2005), Quebec (494 to 806), Ontario (682 to 791), and British Columbia (361 to 999). Alberta was not included in the statistics because it is the only Canadian province that had a privatized retail system, although it continued to retain its wholesale monopoly. The marked progression of the physical availability of alcohol since the 1990s is expected to have led to an increase in alcohol consumption among Canadians, with variations by province. This hypothesis is supported by a systematic review of the literature performed by Popova, Giesbrecht, Bekmuradov, and Patra (2009), which showed that outlet density and hours and days of sale were associated with overall consumption, drinking patterns, or damage caused by alcohol consumption.
Media Coverage of the Protective Effects of Alcohol. The third contextual change concerns the media coverage of the protective effects of moderate drinking. Since the 1970s, health promotion efforts targeting healthy lifestyles have been the focus of public health authorities in Canada (Brisson, Reference Brisson2010). A study by St. Leger, Cochrane, and Moore (1979) found a negative association between alcohol consumption and ischemic heart-disease deaths in countries where moderate wine drinking was frequent (1–3 drinks a day). French epidemiologist and nutritionist Renaud then “discovered” the French Paradox: the French had the lowest mortality rates from heart disease despite higher risk from higher levels of cholesterol, fat, and hypertension due to a high intake of saturated fat (Renaud & de Lorgeril, Reference Renaud and de Lorgeril1992). In addition, an American epidemiological study by Klatsky, Armstrong, and Friedman (Reference Klatsky, Armstrong and Friedman1992) showed a reduced risk of cardiovascular mortality among light drinkers, especially among older persons. It is now well-known that a healthy lifestyle includes regular physical activity, low stress levels, and a healthy diet, such as the traditional Mediterranean diet, which includes drinking moderate amounts of alcohol with meals (Public Health Agency of Canada, 2011).
Several literature reviews and meta-analyses on the protective effects of alcohol were published during the 1990s. Moderate alcohol consumption is associated with a decreased risk of cardiovascular morbidity and mortality and of all-cause mortality (Bagnardi, Zambon, Quatto, & Corrao, Reference Bagnardi, Zambon, Quatto and Corrao2004; Di Castelnuovo et al., Reference Di Castelnuovo, Costanzo, Bagnardi, Donati, Iacoviello and De Gaetano2006; Ronksley, Brien, Turner, Mukamal, & Ghali, Reference Ronksley, Brien, Turner, Mukamal and Ghali2011). Studies indicate that this association is defined by a J-shaped curve that shows abstainers and heavy drinkers to be at increased risk of mortality compared to moderate drinkers. In addition, men over the age of 40 years and post-menopausal women are more likely to experience the benefits of moderate and regular drinking (Dixon et al., Reference Dixon, Dixon and O’Brien2003; Ellison, Reference Ellison2002).
The data received widespread media attention, especially among older cohorts in the population. A press review of Canadian print media (newspapers and journals) conducted by the lead author at Éduc’alcool indicates that between April 2001 and April 2005, approximately 60 articles on moderate alcohol consumption or the positive effects of alcohol were published in Canada. The review did not include, however, electronic media (radio, television, and the Internet). The results of two Canadian studies have also shown public opinion to be predominantly positive towards alcohol and its effects. In 1996–1997, 57 per cent of Canadians thought that moderate alcohol consumption had beneficial effects on health (Ogborne & Smart, Reference Ogborne and Smart2001), and 88 per cent of Ontarians reported that drinking one to two glasses of wine a day could reduce the risk of developing heart disease (Smart & Ogborne, Reference Smart and Ogborne2002).
In sum, the development of alcohol marketing, the increased physical availability of alcohol, and media coverage of the health benefits of moderate drinking have together created a context that is more favourable to alcohol consumption among the Canadian population and especially among adults aged 55 or older. As stated by Giesbrecht (Reference Giesbrecht2007), with the significant promotion of alcohol and normalization of its use into everyday life, Canadian public opinion may tend to see alcohol as a more “benign” product. This surely has an impact on drinking habits.
Alcohol Sales and Consumption. Alcohol sales and consumption per capita have increased in Canada since the 1990s, which may be explained by the aforementioned contextual changes. Given that alcohol sales reflect consumption in a population, greater sales should indicate greater consumption. More specifically, alcohol consumption per capita increased by almost 7 per cent in Canada between 1994 and 2004, from 7.4 to 7.9 litres of absolute alcohol per capita among those 15 years of age or older (data derived from sales of alcoholic beverages, Statistics Canada, 1997d, 2005). The Canadian data show an increase in sales of alcoholic beverages and in alcohol consumption per capita between 1994 and 2004, with provincial variations and differences in popularity for the various types of alcoholic beverages (Statistics Canada, 1997d, 2005).
Between 1994–1995 and 2004–2005, the value of alcoholic beverage sales in Canada increased annually by an average of 4.9 per cent, increasing from $10.4 billion in 1994–1995 to $16.8 billion in 2004–2005. In 1994–1995, sales of spirits, beer, and wine represented 28.3 per cent, 53.0 per cent, and 18.8 per cent, respectively, of all alcoholic beverage sales, whereas in 2004–2005 they accounted for 24.3 per cent, 50.4 per cent, and 25.2 per cent. Thus, beer continued to be the most popular beverage, but wine sales exceeded spirits sales for the first time in 2004–2005. This trend may well be reflected in the drinking behaviours of the Canadian population between the 1990s and the 2000s, with provincial variations. This hypothesis applies to both the general population and to older adults.
Socio-demographic Changes among Older Adults
In addition to the contextual changes, the socio-demographic changes among Canadians aged 55 and older suggest changes in their alcohol consumption between the 1990s and the 2000s. Statistics Canada data indicate that the level of education (Statistics Canada, 1997c, 2008b) and average income (Statistics Canada, 1997a, 2008c) of older Canadian men increased substantially between 1996 and 2006, whereas the marital status (Statistics Canada, 1997b, 2007) and labour force activity (Statistics Canada, 1997c, 2008a) remained relatively stable. More specifically, the great majority of older men are married/partnered: 77 per cent in both 1996 (60 and older) and 2006 (55 and older). Those aged 45–64 are predominantly employed in the labour force (72% in 1996 and 76% in 2006), and the majority of men aged 65 or older are not in the labour force (87% in 1996 and 84% in 2006). Data on education show that the proportion of men having no degree, certificate, or diploma decreased from 36 per cent to 19 per cent during this period among those 45–64 years of age and from 61 per cent to 39 per cent among those aged 65 and older. In addition, the average income of men 55–64 years of age increased between 1990 and 2005 (from $46,114 to $53,486), while that of men 65 and older remained relatively stable over those 15 years ($36,023 to $37,659). These changes are expected to be associated with an increase in alcohol consumption (Abdel-Ghany & Silver, Reference Abdel-Ghany and Silver1998; Demers & Poulin, Reference Demers, Poulin, Adlaf, Begin and Sawka2005; Greenfield, Midanik, & Rogers, Reference Greenfield, Midanik and Rogers2000).
Among older Canadian women, socio-demographic changes are most apparent with regard to marital status (Statistics Canada, 1997b, 2007), the increase in the level of education (Statistics Canada, 1997c, 2008b), and participation in the labour force (Statistics Canada, 1997c, 2008a). Between 1996 and 2006, the data show a decrease in the proportion of women who were married/partnered (from 78% to 55%) and a decrease in those who were single/never married (from 11% to 6%), whereas the proportion of divorced/separated/widowed women rose from 11 per cent to 39 per cent. However, the age groups are not the same for the purposes of marital status; the older adult category was 60 years of age or older in 1996 and 55 or older in 2006. The level of education also increased with time, as the proportion of women in the 45–64 age group who had no degree, certificate, or diploma declined from 40 per cent in 1996 to 19 per cent in 2006. Among women aged 65 and older, the proportion with no degree, certificate, or diploma declined from 66 per cent in 1996 to 46 per cent in 2006. The proportion of women 45–64 years of age in the labour force rose from 60 per cent in 1996 to 68 per cent in 2006. In both 1996 and 2006, almost all women aged 65 and older were not in the labour force (95% and 93% respectively). The observed changes among women are also consistent with an increase in alcohol consumption.
Given the contextual and socio-demographic changes in Canada since the 1990s as well as the increase in alcohol consumption per capita and sales, adults aged 55 years or older in the 2000s should present different rates of drinkers and alcohol consumption profiles compared with those from the previous decade. One would expect to find a new generation of drinkers among the older adults of the 21st century.
Objectives and Hypotheses
Objective 1. The contextual changes associated with alcohol consumption since the 1990s have created a context that is more favourable to alcohol consumption in Canada. The first objective of the study was to determine whether alcohol consumption among Canadians aged 55 or older differed between the 1990s and the 2000s. Older Canadians were expected to display higher rates of drinkers but lower rates of binge drinking in the 2000s. In view of the data on alcoholic beverage sales in Canada, they should also have displayed higher rates of wine and beer drinkers in the 2000s but lower rates of spirits drinkers. Drinkers should also have drunk more frequently in the 2000s.
Objective 2. It is impossible to verify empirically whether contextual (time period) changes explain the changes in alcohol consumption between the 1990s and the 2000s. However, it is possible to measure key socio-demographic indicators. The second objective was thus to verify whether the differences in alcohol consumption between the 1990s and the 2000s persisted when socio-demographic indicators were taken into account. If socio-demographic indicators were controlled for and differences persisted, it could be deduced that the contextual changes were partly responsible for the differences we found between the 1990s and the 2000s.
Men and women were analysed separately and not compared in this study. Differences between men and women were well documented in regard to alcohol consumption. In all countries and during all historical periods for which population surveys have been conducted, men have been over-represented among drinkers, have been shown to drink more frequently and in greater quantities, and have displayed more alcohol-related problems than women (Wilsnack & Wilsnack, Reference Wilsnack and Wilsnack2002). The research hypotheses for the two genders were the same because the contextual and socio-demographic changes were expected to affect both Canadian men and women.
Methods
Participants and Procedure
Data were drawn from two cross-sectional Canadian surveys: the CADS (Canada’s Alcohol and Other Drugs Survey, Statistics Canada, 1994) and the GENACIS survey (GENder, Alcohol, and Culture: An International Study, Graham et al., Reference Graham, Demers, Nadeau, Rehm, Poulin and Dell2003), conducted in 1994 and 2004 respectively. A random digit dialing (RDD) telephone sampling method across the 10 Canadian provinces was used in both surveys. Interviews were completed with one randomly selected household member aged 15 years or older for the CADS and 18 or older for the GENACIS survey. Respondents were interviewed in the official language of their choice (French or English) using computer-assisted telephone interviewing (CATI). The overall response rate was 75.6 per cent (n = 12,155) for CADS and 52.9 per cent (n = 14,067) for GENACIS. This study included subsamples of respondents aged 55 to 74 consisting of 1,071 men and 1,446 women from CADS and 1,494 men and 2,176 women from GENACIS. Only a small number of respondents were aged 75 or older, so we decided not to include them in the analyses.
Measures
Dependent Variables. We used several measures to assess alcohol consumption in both surveys. Rates of current drinkers include those who consumed any alcoholic beverages during the 12 months prior to the survey (CADS: “During the past 12 months, have you had a drink of alcohol?” and GENACIS: “Did you consume any alcohol beverages such as wine, beer, hard liquor, sherry, coolers, or any other beverages containing alcohol during the last 12 months?”).
The frequency of consuming five drinks or more on a single occasion (binge drinking) was a continuous variable in the CADS but categorical in the GENACIS, making comparison between surveys almost impossible (CADS: “How many times in the past 12 months have you had 5 or more drinks on one occasion?” and GENACIS: “How often during the past 12 months have you consumed 5 or more drinks in a single day?: never, less than once a month, 1–3 days a month, 1–2 days a week, 3–4 days a week, every day or nearly every day?”). We thus decided to recode binge drinking into a three-category variable: (1) never (CADS = 0 and GENACIS = never), (2) less than monthly (CADS < 12 and GENACIS = less than once a month), and (3) monthly or more often (CADS ≥ 12 and GENACIS = 1–3 days a month, 1–2 days a week, 3–4 days a week, every day or nearly every day). Because the number of male and female drinkers reporting binge drinking monthly or more often was too small in both surveys to conduct further analyses (as Table 2 shows), we decided to create a dummy (binary) variable discriminating between those who reported at least one occasion of binge drinking over the previous year (1) from those who did not (0).
With regard to beverage preferences, both surveys asked respondents about wine, beer, and spirits consumption over the past 12 months (CADS: “What type of alcohol do you drink? [each type of beverages]: yes/no” and GENACIS: “Do you drink [each type of beverages]?: yes/no”). Thus, drinkers of wine (wine, port, sherry, and/or vermouth drinkers), beer (including light beer drinkers), and spirits (such as vodka, rye, rum, scotch, gin, and/or liqueur drinkers) include all those who had consumed any of those types of alcoholic beverages in the previous year (not mutually exclusive categories).
The frequency of drinking was initially a categorical variable in both surveys. However, three categories differed between surveys, again making comparison almost impossible (CADS: “During the past 12 months, on average, how often did you drink alcohol?: less than once a month, 1–3 times a month, once a week, 2–3 times a week, 4–6 times a week, every day?” and GENACIS: “During the last 12 months, how often did you usually have any kind of drink containing alcohol?: less than once a month, 1–3 days a month, 1–2 days a week, 3–4 days a week, 5–6 days a week, every day?”). We thus decided to recode the frequency of drinking into the number of drinking days per year (a continuous variable). In both surveys, less than once a month was recoded into six days per year (1/2 times × 12 months per year), 1–3 times a month was recoded into 24 days per year (2 times × 12 months per year), once a week for CADS and 1–2 days a week for GENACIS were recoded into 52 days per year (1 time × 52 weeks per year), 4–6 times a week for CADS and 5–6 days a week for GENACIS were recoded into 260 days per year (5 times × 52 weeks per year), and every day was recoded into 365 days per year (1 time × 365 days per year). In the CADS, 2–3 times a week was recoded into 156 days per year (3 times × 52 weeks per year), and in GENACIS, 3–4 days a week was recoded into 208 days per year (4 days × 52 weeks per year).
The usual quantity of drinking per drinking day is a continuous variable indicating the number of drinks the respondent usually had on drinking days over the previous 12 months (CADS: “During the past 12 months, on the days that you had a drink, how many drinks did you usually have?” and GENACIS: “In the past 12 months, on those days when you had any kind of beverage containing alcohol, how many drinks did you usually have?”).
The maximum quantity of alcohol consumed is a continuous variable referring to the highest number of drinks the respondent had consumed on any one occasion over the previous 12 months (CADS: “In the past 12 months, what is the highest number of drinks you can recall having on any one occasion?” and GENACIS: “During the last 12 months, what is the largest number of drinks you had on any single day? Please try to remember the most drinks you had, whether it was five, ten, twenty, twenty-four or more.”).
We also created a continuous variable indicating the annual volume of drinking (number of drinks per year). It was derived from the product of the annual frequency of drinking and the usual quantity of consumption per drinking day.
Independent Variables. The independent variables are socio-demographic indicators, such as marital status, living arrangement, level of education, and employment status. The income measure was excluded because the number of missing values was disproportionately high. Marital status is a three-category variable: married/partnered, divorced/separated/widowed, and single/never married. Living arrangement is a binary variable: 1 (living alone) or 0 (not living alone). Level of education is a four-category variable ranging from 1 (less than secondary) to 4 (university degree). Employment status consists of three categories: working for pay, retired, and other. The “other” category refers to those going to school, caring for family, on disability, or unemployed. In addition, the year of the survey (0 = 1994 and 1 = 2004) is used as a proxy for the time period effect (i.e., the effect of contextual changes).
Control Variables. Control variables include province of residence, age, and subjective health. The province of residence is used as a control variable because alcohol consumption and beverage preferences tend to differ across provinces (Paradis, Demers, & Picard, Reference Paradis, Demers and Picard2010; Statistics Canada, 1997d, 2005). In both surveys, Canadian provinces were recoded into five categories to assure an appropriate number of respondents in each provincial category: Atlantic Provinces (New Brunswick, Nova Scotia, Prince Edward Island, Newfoundland and Labrador), Quebec, Ontario, British Columbia, and Western Provinces (Alberta, Saskatchewan, Manitoba). Because of a possible age/maturation effect, we included age and subjective health as control variables. Age is a two-category variable: 55–64 and 65–74 years of age. We had to use age as a categorical variable because the CADS database only includes age categories. Subjective health is a categorical variable assessing how respondents perceived their physical health as compared to others their age (CADS and GENACIS: “In general, compared to others your age, how has your physical health been in the last 12 months: (1) excellent, (2) very good, (3) good, (4) fair, or (5) poor?”). Previous research has shown a strong association between the subjective health measure and other direct and indirect health measures, such as health assessments performed by general physicians (Bergner, Bobbitt, Pollard, Martin, & Gilson, Reference Bergner, Bobbitt, Pollard, Martin and Gilson1976; Brazier et al., Reference Brazier, Harper, Jones, O’Cathain, Thomas and Usherwood1992; Idler & Angel, Reference Idler and Angel1990; Laforge, Williams, Harford, & Parker, Reference Laforge, Williams, Harford and Parker1990). Subjective health is therefore used as a proxy for actual health status. As an ordinal variable having more than four categories, subjective health was used as a continuous variable in the analyses.
Statistical Analyses
In view of the study’s first objective, univariate analyses of variance (one-way ANOVAs) were conducted to determine whether the alcohol consumption of Canadians aged 55 to 74 differed between 1994 and 2004. Thus, each drinking variable (each dependent variable) was compared between 1994 and 2004. Note that ANOVA was used to compare rates as they refer to 0–1 binary variables.
To achieve the study’s second objective, we performed sequential multiple regression analyses for each alcohol-related variable presenting significant differences between 1994 and 2004 (in this case, logistic regressions for binary dependent variables; see results section). The aim was to assess the effect of time period by verifying whether the differences between 1994 and 2004 persisted when adjusting for control and socio-demographic variables. Thus, we first entered the year of the survey, provinces, age groups, and subjective health (control variables) into the regression model (Model 1). The socio-demographic variables were then entered into the regression model to determine whether the survey year effect was modified (Model 2). The survey year effect was assessed using odds ratios (OR) with Wald tests and associated confidence intervals (95% CIs). All analyses were performed separately by gender using Predictive Analytics Software (PASW) Statistics 18.
Results
Descriptive Characteristics of Subsamples
Table 1 shows the descriptive characteristics of men and women aged 55–74 in 1994 and 2004 (socio-demographic and control variables). The majority of men were married/partnered in both 1994 and 2004 (71.2% and 72.5% respectively), approximately 20 per cent were divorced/separated/widowed, and approximately 8 per cent were single/never married. About one-quarter of the men lived alone in both 1994 and 2004. The proportion of men with some post-secondary education (including university degree) increased from 30.9 per cent in 1994 to 46.7 per cent in 2004. Employment status shows an increase in men working for pay from 34.1 per cent in 1994 to 45.6 per cent in 2004, whereas the proportion of retired men dropped from 58.6 per cent in 1994 to 47.2 per cent in 2004. Approximately 7 per cent of men in both 1994 and 2004 had an “other” employment status.
Table 1: Descriptive characteristics of Canadians aged 55–74, by gender and year of survey
In 1994, 49.3 per cent of women were married/partnered, which increased to 56.9 per cent in 2004. The proportion of divorced/separated/widowed women dropped from 44.5 per cent in 1994 to 35.8 per cent in 2004. A similar proportion of women were single/never married in 1994 (6.2%) and in 2004 (7.3%). The proportion of women living alone decreased from 45 per cent in 1994 to 35.3 per cent in 2004. In addition, the proportion of women having some post-secondary education (including university degree) rose from 29.4 per cent in 1994 to 43.5 per cent in 2004. In 1994, 18 per cent of women worked for pay, which increased to 31.4 per cent in 2004, and 31 per cent of women in 1994 reported “other” as their employment status compared with only 13.9 per cent in 2004. A similar proportion were retired in 1994 (51%) and 2004 (54.6%).
The control variables indicate similar or slightly different (Atlantic Provinces) provincial repartitioning between samples, but the proportions of men and women in each age group differ slightly. Additionally, about half the men and women in both 1994 and 2004 perceived themselves as being in excellent or very good physical health.
Were There Differences in Alcohol Consumption between 1994 and 2004?
Among older men and women, there were differences between 1994 and 2004 in rates of drinkers (how many) but not drinking profiles (how). More specifically, Table 2 shows that rates of current drinkers were significantly higher in 2004 than in 1994 (men: 74.3% vs. 67.1% respectively; women: 65.9% vs. 51.8% respectively). Similarly, rates of binge drinking were higher in 2004 than in 1994 (men: 44.1% vs. 33.4% respectively; women: 15.7% vs. 8.8% respectively). Rates of wine drinkers were significantly higher in 2004 (men: 75.7% vs. 60.6% respectively; women: 85.2% vs. 80.5% respectively). Rates of beer drinkers were significantly higher in 2004 among men (77.6% vs. 69.8% respectively), whereas no significant differences were found among women. Rates of spirits drinkers were significantly higher in 1994 among both men (65.8% vs. 60.5% respectively) and women (53.1% vs. 48.3% respectively). The results show no significant differences between 1994 and 2004 for either men or women in average annual volume of drinking, average annual frequency, average usual quantity, or average maximum quantity.
Table 2: Comparison of rates of drinkers and drinking profiles among Canadians aged 55–74 between 1994 and 2004, by gender
* p < .05; ** p < .01; *** p < .001
SD = standard deviation
a Missing responses are taken into account in reported rates
b Among drinkers exclusively
Do the Differences Persist When Controlling for Socio-demographic Indicators?
Results of the sequential logistic regression analyses are presented in Table 3 for men and in Table 4 for women. As noted, we conducted a regression analysis for each drinking variable that presented significant differences between 1994 and 2004 according to the ANOVAs performed previously (men: rates of current, binge, wine, beer, and spirits drinkers; women: rates of current, binge, wine, and spirits drinkers).
Table 3: Rates of male drinkers aged 55–74: the year of the survey effect adjusted for control (model 1) and socio-demographic variables (model 2)
* p < .05; ** p < .01; *** p < .001
a Referent category = Atlantic Provinces
b Referent category = married/partnered
c Referent category = less than secondary
d Referent category = working for pay
CI = confidence interval
OR = odds ratios
R 2= Nagelkerke
Subjective health: 1 = excellent to 5 = poor
Table 4: Rates of female drinkers aged 55–74: the year of the survey effect adjusted for control (model 1) and socio-demographic variables (model 2)
* p < .05; ** p < .01; *** p < .001
a Referent category = Atlantic Provinces
b Referent category = married/partnered
c Referent category = less than secondary
d Referent category = working for pay
CI = confidence interval
OR = odds ratios
R 2= Nagelkerke
Subjective health: 1 = excellent to 5 = poor
The differences between 1994 and 2004 in rates of current drinkers and rates of spirits drinkers among men were not significant after including the control variables (model 1), and they remained non-significant when controlling for socio-demographic variables (model 2; R 2 not significant for rates of spirits drinkers). Conversely, rates of binge drinking (model 1: 1.43, p < .001; model 2: 1.52, p < .001) and of wine (model 1: 1.72, p < .001; model 2: 1.45, p < .01), and beer drinkers (model 1: 1.37, p < .01; model 2: 1.45, p < .01, however R 2is not significant) remained significantly higher in 2004 than in 1994 when adjusted for both the control and socio-demographic variables.
In women, the differences between 1994 and 2004 in rates of wine and spirits drinkers were no longer significant when adjusted for the control variables (model 1), and they remained non-significant when controlling for socio-demographic variables (model 2; R 2 not significant for rates of spirits drinkers). Rates of current drinkers (model 1: 1.60, p < .001; model 2: 1.35, p < .001) and binge drinking (model 1: 1.65, p < .01; model 2: 1.75, p < .001; however, R 2is not significant) continued to be significantly higher in 2004 than in 1994 when adjusted for both control and socio-demographic variables.
For both men and women, in no case did the introduction of socio-demographic variables into the regression models cancel the differences between 1994 and 2004 (for each drinking variable). In some cases, the R-squared value (R 2) associated with the second model was not significant, meaning that the inclusion of the socio-demographic variables in the regression model did not significantly help in “predicting” the dependent variable. When the second models were significant, higher education stood out as an important variable associated with higher rates of current and wine drinkers among men, whereas lower education was associated with higher rates of binge drinkers. Among women, higher education and marital status (being married/partnered) were significantly associated with higher rates of current and wine drinkers.
Discussion
Canadian contextual changes associated with alcohol as well as socio-demographic changes among older adults since the 1990s suggested an increase in alcohol consumption among this subpopulation in the 2000s. However, normal physiological changes with aging generally encourage a decrease in alcohol consumption with age, which would mean that older adults should drink the same way regardless of the time period to which they belong. With this in mind, the present study: (a) compared alcohol consumption in Canadian men and women aged 55 to 74 between 1994 and 2004, and (b) explored whether period differences persisted when controlling for the socio-demographic composition of the samples.
As predicted, rates of drinkers were significantly higher in 2004 than in 1994 among both men and women, and, contrary to the hypotheses, this was also the case for binge drinking. In addition, rates of male and female wine drinkers and rates of male beer drinkers were higher in 2004, while rates of spirits drinkers were higher in 1994 among both men and women. The results also showed that the socio-demographic differences between 1994 and 2004 did not cancel the differences in the various types of rates of drinkers. Additionally, contrary to the hypotheses, rates of beer drinkers among women did not differ between 1994 and 2004: at both measurement times, the proportion of beer drinkers was particularly low (approximately 35%). These results indicate the well-known female preference for wine rather than beer (Paradis et al., Reference Paradis, Demers and Picard2010). Contrary to expectations, alcohol consumption profiles did not differ between 1994 and 2004 for either gender: average volume, frequency, quantity, and maximum quantity of drinking were not significantly different.
It must be concluded that socio-demographic differences do not explain the differences observed in this study between 1994 and 2004. Including age group in the regression analyses restricted the cohort effect, such that the remaining effect was mostly due to time period changes. Contextual changes that occurred since the 1990s are thus potentially responsible for the increases in the rates of different types of drinkers between 1994 and 2004. Media coverage of the benefits of moderate alcohol consumption among older adults represents a major shift in perception of the effects of alcohol in late life. Drinking may not be perceived as a particularly risky behaviour, and the higher rates of current drinkers in the 2000s may be explained by older adults aiming at having healthy lifestyle habits and thus not transitioning to abstinence (Shaw, Krause, Liang, & McGreever, Reference Shaw, Krause, Liang and McGreever2011). In addition, the fact that older adults seemed to have replaced spirits with wine and beer in 2004 suggests that the positive messages regarding the benefits of the Mediterranean diet may be effective.
Proportionally more men and women from 2004 had binge drunk over the past year compared with those in 1994, but the frequency of binge drinking in both 1994 and 2004 was typically less than once a month. This means that, in all probability, anyone attending major celebrations (i.e., a New Year’s Eve party or a wedding) ended up in that category. This drives us to conclude that the higher rates of binge drinking may be related to the higher rates of drinkers; this is likely not alarming from a public health perspective because the frequency seems fairly low. This could also explain why we observed higher rates of binge drinking in 2004 but no significant differences between 1994 and 2004 in drinking profiles (volume, frequency, quantity, maximum quantity). More data are needed on binge drinking among older Canadian adults to understand better the situation and determine whether the present results are typical or atypical.
The increased physical availability of alcohol in Canada since the 1990s may also explain the increase in the different rates in terms of types of drinkers between 1994 and 2004 among those aged 55 to 74. In keeping with the premises of the alcohol availability theory (Bruun et al., Reference Bruun, Edwards, Lumio, Makela, Pan and Popham1975), our results indicate that (1) increased availability of alcohol between 1994 and 2004 may be associated with higher rates of drinkers among adults aged 55–74, and (2) higher rates of drinkers are potentially associated with higher rates of binge drinking, although binge drinking was not a frequent behaviour in our subsamples. The third proposition of the alcohol availability theory – that a greater number of binge drinkers should be associated with a greater number of alcohol-related problems – remains to be verified, as it is not possible to do so with the current data. Frequent binge drinking may cause more alcohol-related problems than binge drinking less than once a month.
This study also showed that differences between 1994 and 2004 in rates of current drinkers among men, of wine drinkers among women, and of spirits drinkers among both genders were cancelled by the introduction of the control variables (provinces, age groups, and subjective health) into the regression models. It would be interesting to investigate how each of these specific variables is related to drinking habits among older adults. Given the different variations in alcohol consumption associated with the various Canadian provinces (Paradis et al., Reference Paradis, Demers and Picard2010), future studies could explore provincial alcohol-related differences in older Canadian adults. Further research should also explore drinking rates and profiles associated with health status according to the different age groups included in the large group that includes everyone aged 55 and older.
This study also revealed that alcohol consumption profiles did not differ between 1994 and 2004. It would therefore seem that neither the contextual changes since the 1990s nor socio-demographic changes affected the alcohol consumption profiles in this age group. It is equally possible that the contextual and socio-demographic changes acted in opposite ways and that their respective effects cancel each other in the analyses. An alternative explanation is that the normal physiological changes with aging are responsible, at least in part, for the similar consumption profiles among older adults at the two measurement times. Alcohol consumption may not rise for older adults over time. Further analyses are necessary to explore age-cohort-period effects on alcohol consumption among Canadian older adults. Past longitudinal analyses of the American general population performed by Moore et al. (Reference Moore, Gould, Reuben, Greendale, Carter and Zhou2005) showed that drinking declined with age even after controlling for cohort and period effects. However, alcohol consumption declined more slowly with increasing age among recent cohorts than among earlier cohorts. Similarly, Glynn et al. (Reference Glynn, Bouchard, LoCastro and Laird1985) studied age and cohort effects using two time points in the Normative Aging Study and found a decrease in alcohol consumption over time, with earlier cohorts drinking less than more recent cohorts. Levenson, Aldwin, and Spiro (Reference Levenson, Aldwin and Spiro1997) analysed three time points from the Normative Aging Study and found a tendency for drinking to decrease with age, although it varied by period and cohort. The period effect was particularly strong in their analyses.
Similarly, Canadian data indicate that among both men and women aged 15 or older, the percentage of heavy drinking (14 alcoholic drinks or more per week for men and 12 or more for women) has decreased by more than half between 1978–1979 and 2000–2001 (Wister, Reference Wister2005). Epidemiological data show a smoothing in heavy drinking rates across age groups over time (approximately 10% among men and between 2% and 4% among women in 2000–2001), pointing at small age effects as compared with cohort and period effects. The shift in heavy drinking patterns may reflect health promotion efforts and changes in social norms regarding drinking (Wister, Reference Wister2005). The results of the present study thus partly corroborate past research (i.e., potential period effect) but call for more elaborated longitudinal analyses of drinking among Canadian older men and women.
Limitations
The main limitations of the study include the cross-sectional CADS and GENACIS data, which do not allow for observation of changes in alcohol consumption among adults aged 55–74 between 1994 and 2004 but make it possible to observe differences between the two points in time instead. It is impossible to draw causal inferences related to the significant associations reported in this study. Our analyses would have been stronger if we had used other cross-sectional survey points rather than only two. The 10-year period is too brief to identify historical changes, thus (a) broader societal period effects are possibly underestimated, and (b) the period under observation could have been atypical. Furthermore, individuals 55 years old in 1994 may have been part of the 65–74 age group in 2004, and thus the samples may not be entirely independent. However, the inclusion of the age group as a control variable in the analyses reduced the potential effect of this limitation. In addition, with increased life expectancy in Canada and the reversal of the age pyramid, it will be crucial to include adults aged 75 and older in future samples to learn more about their relationship to alcohol.
Another limitation is that the data were self-reported and may include recall and reporting biases. Memory errors and potential stigmata associated with alcohol consumption may lead to an underestimation of drinking (Graham, Reference Graham1986). One of the study’s major limitations concerns the transformation of drinking frequency into days per year because the categories differed between the two surveys. The results regarding drinking frequency and volume (which is derived from the frequency) must therefore be interpreted with caution.
Conclusion
The increase in the marketing and physical availability of alcohol as well as in media coverage of the health benefits of drinking since the 1990s may have contributed to the increase in rates of current drinkers and to the replacement of spirits with wine and beer drinking among older Canadian adults in the 2000s, indicating potentially healthier changes in the relationship with alcohol in older adulthood. Overall consumption levels did not shift, and the mean frequency and quantity of drinking reflect healthy (moderate) drinking habits among our subsamples. More research is needed to disentangle the age-cohort-period effects on alcohol consumption among Canadian older adults.
