Abstract
The relationship between adverse life events, including conjugal bereavement, and alcohol consumption among older people is not well established. Many older people experience significant psychosocial stressors, but only a small minority develops substance abuse [1]. While clinical data support the contention that alcohol consumption among lifelong drinkers may increase following adverse life events [2–4], a cross-sectional survey among a large, non-clinical, sample failed to find a similar relationship [5]. More recently, Welte and Mirand found that among 2325 New York residents ‘stress’, as measured in a variety of ways, was not related to heavy drinking [6].
However, it is well established that conjugal bereavement in older people leads to considerable emotional distress [7] and is associated with an increased risk of depressive episodes [8] and suicide [9]. For older men, there is ample evidence also that conjugal bereavement is associated with excess mortality [10,11]. While this excess mortality has been found to be due to diverse ‘natural causes’ [12], its precise aetiology remains obscure. One hypothesis is that recently widowed older men become emotionally distressed following the death of their spouse and, as a result, increase certain prejudicial health-related behaviours, including alcohol consumption, which lead ultimately to excess mortality. An alternative hypothesis is that the death of their spouse leads to loss of care and control over putative prejudicial health-related behaviours. In the study reported here, we investigated the first of these hypotheses in relation to alcohol consumption. The main specific hypothesis we tested was that widowers' alcohol consumption is predicted by their level of self-reported psychological distress. A subsidiary hypothesis was that older widowers have higher alcohol consumption than matched married men. We chose to study widowers rather than widows as most studies of mortality following bereavement have concluded that excess mortality is restricted to men and because the available epidemiological data suggest that men are far more likely to abuse alcohol in later life than women. In addition, the health-related behaviour of older widows has been the subject of many studies whereas recently widowed older men have rarely been the subject of systematic investigation.
Method
Subject recruitment has been described in detail elsewhere [13]. Briefly, 57 recently widowed, community-residing, older men were identified from the official death records of their wives over a 4-month period. A comparison group of 57 married men was also enrolled. The names of up to 20 potential married subjects for each widower were obtained from the electoral roll and matched for sex, age, socioeconomic status and locality of residence. The best matching, community-residing, married man was selected in each case. All subjects were free of clinically significant cognitive impairment. Widowers were interviewed on three occasions: at 6 weeks (T1), 6 months (T2) and 13 months (T3) post-bereavement. Married men were interviewed at similar intervals.
Basic sociodemographic data including occupational prestige [14] were obtained from each subject. Cognitive performance was assessed using the Mini-Mental State Examination (MMSE) [15].
Self-reported current alcohol consumption over the previous month was measured by the quantity/frequency method as used in the Risk Factor Prevalence Survey [16]. Quantity of alcohol consumption was measured on a seven-point ordinal scale: 1, nil; 2, 1–2 drinks/day; 3, 3–4 drinks/day; 4, 5–8 drinks/day; 5, 9–12 drinks/day; 6, 13–20 drinks/day; 7, >20 drinks/day. Frequency of alcohol consumption was measured on a six-point ordinal scale: 1, nil; 2, < 1 day/week; 3, 1–2 days/week; 4, 3–4 days/week; 5, 5–6 days/week; 6, 7 days/week. A standard drink was defined as any alcoholic beverage containing approximately 10 g ethanol. Because of the ready availability of low-alcohol beer, an additional question with a six-point ordinal scale inquired about consumption of low-alcohol beer. The Australian National Health and Medical Research Council (NHMRC) has recommended that adult males consume no more than four standard drinks of alcohol per day [17]. It has further recommended that drinkers have at least two alcohol-free days per week. Thus, to make analysis of the present data meaningful in terms of the NHMRC recommendations, the quantity data were dichotomised around the response categories 3/4. The frequency data were dichotomised around the response categories 4/5, providing an approximation to the NHMRC categories.
To assess the validity of self-reported alcohol consumption, blood samples were taken on two occasions to measure serum liver enzyme levels. Three liver enzymes were measured: gamma glutamyl transferase (GGT), alanine amino transferase (ALT), and aspartate amino transferase (AST). Elevations in the serum levels of these liver enzymes are commonly found in older persons using or abusing alcohol [18].
The Bereavement Phenomenology Questionnaire (BPQ) [19], a 22-item interviewer-administered, structured questionnaire, was used to assess the level of grief phenomena among widowers in the 2 weeks prior to each interview. The BPQ employs a four-point frequency scale from 0, ‘never’, to 3, ‘often’, and has satisfactory validity and reliability [19].
As our previous findings [13] had indicated that state anxiety as measured on the State/Trait Anxiety Inventory (STAI) [20] was the psychological variable which best distinguished widowers from married men, this measure was used to assess the relationship between emotional distress and self-reported alcohol consumption. The 20-item state component of the STAI is a self-report questionnaire that uses a four-point frequency scale from 1, ‘almost always’, to 4, ‘almost never’, to assess current level of anxiety. It has demonstrated validity and reliability [21] and has been employed previously in a study of widowers and married men [22].
Statistical analyses
Correlational analyses between categorical or ordinal data were undertaken using Spearman's rho (rs). Repeated measures categorical data were analysed using the Mantel-Haenszel Chi-squared (MHχ2) test [23]. Dimensional data were checked for normality and differences between means assessed using analysis of variance. The SPSS (SPSS Inc., Chicago, IL, USA) and S-PLUS (Statistical Sciences Inc., Seattle, WA, USA) statistical software programs were used.
Results
Widowers had a mean age of 74.5 years (SD = 4.7), a mean MMSE score of 27.6 (SD = 1.8), and a mean occupational prestige of 4.7 (SD = 1.2). Married men had a mean age of 75.4 years (SD = 6.2), a mean MMSE score of 27.6 (SD = 1.8), and a mean occupational prestige of 4.7 (SD = 1.8). The majority of widowers (70.1%) and married men (66.7%) rated their general physical health as ‘excellent’ or ‘good’. Most widowers (75.0%) and married men (54.5%) had had fewer than 10 years of formal education. There were no significant differences between widowers and married men on any of these variables.
Self-reported alcohol consumption data are summarised in Table 1. Similar proportions of widowers and married men reported drinking some alcohol over the period of the study. Although a higher proportion of widowers (36.5%) than married men (18.2%) reported drinking no alcohol at T3, this difference was not statistically significant (χ2 = 1.98, df = 1, p = 0.159). Each of the 2 × 2 × 3 alcohol consumption data arrays was assessed for statistical significance using the Mantel-Haenszel Chi-squared test (MHχ2). In comparison with married men, widowers reported both greater frequency of alcohol consumption (MHχ2 = 4.64, df = 1, p = 0.031) and greater quantity of alcohol consumption (MHχ2 = 7.75, df = 1, p = 0.0054). When subjects who reported that they mainly consumed low alcohol beer were eliminated from further consideration and the remaining data reanalysed, similar results were found for frequency (MHχ2 = 3.94, df = 1, p = 0.047) and quantity (MHχ2 = 8.79, df = 1, p = 0.003) of alcohol consumption. Thus, it can be concluded that the elderly widowers reported consuming significantly more alcoholic drinks on more days than matched married men.
However, a further consideration arises: could the group differences reflect a response bias with widowers merely reporting greater alcohol intake than married men? This question was tested indirectly by measuring the Spearman rank correlations (rs) between self-reported alcohol consumption and the serum concentrations of three liver enzymes (GGT, ALT and AST) in both widowers and married men on two occasions (T1 and T3). There were highly significant correlations between both reported quantity and reported frequency of alcohol consumption, and the serum concentrations of GGT, ALT and AST at both T1 and T3 among widowers (see Table 2). The correlations between alcohol consumption and serum liver enzyme levels among married men were weaker, with none reaching significance at T3 (see Table 3).
Self-reported frequency and quantity of alcohol consumption in older widowers and married men
Spearman correlations (rs) between serum levels of liver enzymes and self-reported alcohol consumption among widowers
Spearman correlations (rs) between serum levels of liver enzymes and self-reported alcohol consumption among married men
The lower correlations evident among married men were probably because of reduced variance in their alcohol consumption data. In addition, widowers had higher serum levels of these liver enzymes than married men (Table 4), although only for AST did these differences reach statistical significance (at T1: F1,97 = 5.23, p = 0.024; at T3: F1,95 = 5.71, p = 0.019). Together, these findings provide evidence for the validity of the self-reported quantity and frequency data and provide further support for the hypothesis that there is a real difference between the alcohol consumption of recently widowed older men and married older men.
Among widowers, correlational analyses were undertaken between self-reported state anxiety (STAI) and self-reported alcohol frequency and quantity at T1, T2 and T3 (Table 5). Using Spearman's correlation coefficient (rs) and two-tailed tests, there were no significant correlations between STAI state anxiety and alcohol consumption. Similar correlational analyses were undertaken between total scores on the Bereavement Phenomenology Questionnaire (BPQ) and self-reported alcohol frequency and quantity at T1, T2 and T3 (Table 6). There were no significant correlations between BPQ scores and alcohol consumption. Thus, the emotional distress experienced by recently widowed older men, whether measured as bereavement-specific phenomena or as non-bereavement-specific state anxiety, was not related to their self-reported alcohol consumption.
Serum liver enzyme levels among older men at T1 and T3 according to bereavement status
Spearman correlations (rs) between self-reported State Trait Anxiety Inventory state anxiety and alcohol intake in older widowers
Discussion
The data obtained provide evidence of hazardous alcohol consumption among both older widowers and married men. On average, across the three phases of data collection, 18.9% of widowers and 8.3% of married men reported drinking five or more standard alcoholic drinks (equivalent to approximately 50 g of ethanol) per drinking day. This difference is statistically significant. Data from the Australian National Health Survey 1989–1990 show that among men aged 65–75 years, 14.2% reported consuming 40 g or more of alcohol per day, and that for men aged 75++ years, 4.8% reported consuming this amount [24]. In comparison with these data, the widowers in the present study reported higher rates of hazardous alcohol consumption.
Given the evidence of excessive alcohol consumption among widowers, a question arises as to the likely health consequences. While there is considerable debate about likely beneficial cardiovascular effects of moderate alcohol consumption, even in the elderly [25], there is strong evidence that increased alcohol consumption is associated with increased risk of fatal and non-fatal stroke in middle-aged and elderly men [26]. A variety of other health problems are associated with alcohol abuse among older men including falls, motor vehicle accidents, cirrhosis of the liver, suicide and the Wernicke-Korsakoff syndrome [27–29]. Thus, excessive alcohol consumption among recently widowed older men may contribute to their excess ‘all causes’ mortality.
Spearman correlations (rs) between intensity of self-reported grief (Bereavement Phenomenology Questionnaire) and alcohol intake in older widowers
Despite the higher levels of reported alcohol consumption by recently widowed older men, similar proportions of widowers and married men reported that they consumed some alcohol (at T1, 70.2% and 71.9%, respectively). These proportions are somewhat higher than were found in Australian males aged 65++ years in the National Health Survey 1989–1990 in whom the proportion of alcohol drinkers was 60.9% [24]. The reasons for the difference between the rates in the present study and the national data are unclear, although in enrolling recently widowed community-residing older Brisbane men, the present study may have selected a relatively healthy cohort with easy access to alcohol.
Given the finding that similar proportions of widowers and married men reported consuming some alcohol, it seems unlikely that the impact of bereavement and the transition to widowhood had been to convert abstainers to drinkers. However, our data do not allow us to definitively exclude this possibility. What seems more likely to us is that established drinkers might have increased their level of alcohol consumption following the death of their spouse. Although there is some support in the literature for this hypothesis [30,31], it cannot be tested directly from the present data as prior drinking history was not obtained. The main methodological difficulty with obtaining prior drinking history in terms of quantity/frequency data is determining an appropriate reference period prior to bereavement. The period during the spouse's final illness may not be an appropriate comparison period as it may not represent usual drinking behaviour. Many of the widowers in the present study had cared for their wives for lengthy but varying periods during their final illnesses.
The finding of greater alcohol consumption among recently widowed older men is at odds with the findings of Saunders et al., who reported that there was no significant difference between the rates of regular drinking among married men and men who had been widowed 3 years earlier in Liverpool, UK [32]. However, it is possible that the excess alcohol consumption of recently widowed older men diminishes after the first year post-bereavement. Or there may be a more complex relationship between excessive drinking and life events such as conjugal bereavement. Using a longitudinal approach, Glass et al. studied the impact of negative life events on alcohol consumption among 2040 older (65++ years) men and women (mean age = 74 years) [31]. Among men who were married at baseline (n = 474; 59%), subsequent death of spouse (n = 25; 5.3%) did not independently predict change in alcohol consumption in a multiple regression analysis. However, there was a significant interaction effect between death of spouse and baseline alcohol consumption (p = 0.048) with men who consumed greater amounts of alcohol at baseline being more likely to increase their alcohol consumption following conjugal bereavement.
If, as suggested above, the effect of conjugal bereavement on the drinking behaviour of older men is increased consumption of alcohol by established drinkers, one further question arises: is increased alcohol consumption associated with emotional distress? This seems quite unlikely in the present cohort as self-reported alcohol consumption (quantity and frequency) was found to be unrelated to measures of grief and state anxiety.
Regardless of the outcome of the debate on ‘stress’ and alcohol abuse, the high rates of potentially hazardous alcohol consumption among recently widowed older men found in the present study mean that older widowers should be regarded as a high risk group. As such, they would be suitable targets for preventive interventions regarding their alcohol consumption. On the basis of the present findings and those from the literature, it would seem appropriate to base the design of intervention strategies on the premise that heavy drinking among recently widowed older men is likely to be a result of loss of spousal care and control rather than a result of ‘stress’ or emotional distress. Intervention programs for older widowers who are drinking excessively could be specifically designed to assist them to better deal with the antecedents to drinking and/or the consequences of drinking. Intervention programs predicated upon the reduction of emotional distress may be less likely to succeed.
Conclusions
Recently widowed older men were found to be significantly more likely than matched married men to report hazardous alcohol consumption. However, self-reported alcohol consumption was not significantly associated with psychological distress whether measured on a bereavement-specific questionnaire or on a non-bereavement-specific measure of state anxiety. It is argued that older widowers are more likely to have increased their alcohol consumption because of a loss of spousal care and control rather than as a result of emotional distress.
Footnotes
Acknowledgements
The authors would like to thank Neroli Whiteman and Cathy Jong, who provided essential research assistance, and Stuart Bryant, Director, Department of Pathology, Royal Brisbane Hospital, and his staff, who generously supplied venipuncture equipment and laboratory tests. The work was supported, in part by grants from the National Health and Medical Research Council and the Australian Rotary Health Research Fund.