Retired Status and Older Adults’ 10-Year Drinking Trajectories

Abstract

Objective:

Little research has examined the role of retirement in shaping late-life drinking careers, and it has generally been limited to cross-sectional designs or short-term follow-ups that emphasize group-level comparisons of retirees and nonretirees. The purpose of this study was to determine the following: (a) the effect of retired status on older adults’ 10-year within-person drinking trajectories and (b) whether age, gender, income, health, and problem-drinker status account for or moderate this effect.

Method:

We first estimated older adults’ (baseline M= 62 years; n = 595) 10-year within-person drinking trajectories using three successively predictive multilevel regression models: unconditional growth, retired status alone, and retired status controlling for covariates. Next, we determined whether inclusion of Retired Status × Covariate interactions would improve prediction of the trajectories.

Results:

Participants’ drinking frequency declined moderately over the 10-year interval, and retired status hastened the decline. However, this effect disappeared once covariates were added to the model: Baseline poorer health, lower income, and current problem-drinker status predicted steeper decline in drinking frequency, whereas former problem-drinker status predicted slower decline. Lower income and current drinking problems also predicted steeper declines in amount of alcohol consumed. There were no statistically significant or uniquely contributive interactions between retired status and age, gender, health, income, or drinking problems for predicting late-life drinking trajectories.

Conclusions:

Baseline health, income, and problem-drinking history are more important than retired status for predicting older adults’ long-term within-person drinking trajectories. These factors—and recency of drinking problems—should be considered in future studies of retirement and late-life drinking patterns.

Little research has examined the role of retirement in shaping late-life drinking trajectories. Previous cross-sectional (Barnes, 1979) and short-term, two-point assessments (e.g., Bacharach et al., 2004; Ekerdt et al., 1989; Midanik et al., 1995) suggest that there is little or no difference between retirees’ and nonretirees’ patterns of alcohol consumption. However, these studies have been limited to group-level comparisons, which can mask important within-person drinking patterns (Raudenbush, 2001). Further, their time frames may have been too narrow to capture longer term effects of retired status on alcohol consumption patterns. In this regard, Perriera and Sloan (2001) showed that, among older adults followed for 6 years, retired status elevated risk of subsequent increased alcohol consumption.

There also has been limited investigation of demographic and personal characteristics that may account for or moderate the relationship between retired status and late-life drinking trajectories. In general, marital status, race/ethnicity, education, and preretirement job level do not account for or moderate the effects of retirement on changes in later life alcohol consumption (Bacharach et al., 2004; Ekerdt et al., 1989; Gallo et al., 2001). There is limited and conflicting evidence regarding moderating effects of age, gender, health, and income on effects of retirement on later life alcohol consumption (cf. Bacharach et al., 2004; Eckert et al., 1989; Gallo et al., 2001; Perriera and Sloan, 2001; Richman et al., 2006; Satre and Arean, 2005).

Some evidence implicates history of drinking problems as an important influence on retirees’ drinking patterns. Perriera and Sloan (2001) showed that having a history of drinking problems decreased the likelihood that retirees would reduce their alcohol consumption over a 6-year interval. However, they could not distinguish between retirees with active/current drinking problems and those who were former problem drinkers because they assessed drinking history with a single, baseline administration of the CAGE questionnaire, which allows assessment only of whether respondents have “ever” had any of four alcohol-related experiences.

To add to and extend previous research in this area, this study aims to determine the following: (a) the effect of retired status on older adults’ long-term within-person drinking trajectories and (b) whether age, gender, income, health, and current or former problem-drinking status account for or moderate this effect.

Method

Sample

Participants were drawn from a larger sample of community residents who took part in a 10-year, four-wave longitudinal study of late-life drinking behavior (for details, see Brennan and Moos, 1990; Moos et al., 1990). Follow-up rates at 1, 4, and 10 years were 94%, 94%, and 93%, respectively, and most (97%, n = 1,193) of the 10-year surviving participants took part in all four waves of data collection. Of these participants, 443 were consistently unemployed (“retired status”), and 152 were consistently employed at all four study assessments. These 595 participants are the focus of this study.

At baseline, 56% of these participants were male. On average, they were about 62 years old, had an annual income of $40,000, reported 1.6 chronic medical conditions, drank about four times per week, and consumed about 2 oz. of ethanol when they drank. At baseline, 41% of these individuals had one or more current drinking problems, and 23% were former problem drinkers (no current drinking problems but past drinking problems experienced 2 or more years before baseline). Retired participants were significantly more likely to be female, to be slightly older, and to have had somewhat lower income and more medical conditions than did consistently employed participants. However, there were no statistically significant differences between the groups in alcohol consumption or problem-drinker status.

Measures

We assessed participants’ demographic characteristics and their health status (count of 13 chronic medical conditions, diagnosed by a physician, that began more than 1 year before baseline) using items from the Life Stressors and Resources Inventory (Moos, 2002; Moos and Moos, 1994). Alcohol consumption was assessed with items from the Health and Daily Living Form (Moos et al., 1992). Frequency of alcohol consumption was calculated by summing responses to three questions, scored on 5-point scales, that asked how often during a week (never to nearly every day) participants had typically consumed wine, beer, and distilled spirits during the past year. Quantity of alcohol consumed was assessed by first determining the number of drinks within each of three beverage categories (wine, beer, and distilled spirits) participants had consumed on typical drinking occasions during the last year, then converting number of drinks to ounces of ethanol and summing ounces of ethanol across beverages. Current and former problem-drinker status was identified using items from the Drinking Problem Index (Finney et al., 1991; see also Allen and Wilson, 2003), which has high internal consistency and good construct validity (Bamberger et al., 2006; Brennan and Moos, 1990; Finney et al., 1991; Kopera-Frye et al., 1999).

Summary of analyses

We estimated 10-year trajectories of drinking frequency and quantity of alcohol consumed using three successively predictive multilevel regression models: (a) unconditional growth; (b) retired status alone; and (c) retired status controlling for the covariates of age, gender, health, income, and current and former problem-drinker status. We next determined whether there were statistically significant Retired Status × Covariate interactions and whether their inclusion in the third model improved prediction of the 10-year drinking trajectories.

Results

The intercept for the unconditional growth model (Table 1, Model A) indicates that, on average, sample members were drinking four to five times a week at baseline, and the frequency with which they consumed alcohol declined significantly thereafter (i.e., intercept of the slope = −0.22). As shown in Model B, retired participants experienced a significantly steeper decline in frequency of alcohol consumption than did participants who continued to work (slope = −0.21); addition of retired status to the model resulted in significantly improved goodness of fit. However, after age, gender, health status, income, and current and former problem-drinker status were statistically controlled (Model C), no significant effect of retired status on drinking trajectories remained. Independent of retired status, having lower baseline income was associated with less frequent initial alcohol consumption (intercept = −0.14) and a slightly faster rate of decline in drinking frequency (slope = −0.04). Similarly, poorer health at baseline was associated with a steeper decline in drinking frequency (slope = −0.05).

Table 1.

Effects of retired status on level and change in frequency and quantity of alcohol consumption

	Frequency of alcohol consumption			Quantity of alcohol consumed
	Model A	Model B	Model C	Model A	Model B	Model C
Model	Unconditional growth	Retired status alone	Retired status, adjusted for covariates	Unconditional growth	Retired status alone	Retired status, adjusted for covariates
Intercept
Intercept	4.6**	4.7**	1.5	2.4**	2.24**	2.5**
Retired		−0.13	0.14	—	0.27	0.29
Age	—	—	0.04	—	—	0.00
Gender	—	—	−0.53	—	—	−0.76**
Poorer health	—	—	−0.04	—	—	−0.03
Lower income	—	—	−0.14*	—	—	0.05
Problem-drinking history
Current problem drinder	—	—	1.7**	—	—	1.5**
Former problem drinker	—	—	−1.5**	−	−	−0.49**
Slope
Intercept	−0.22**	−0.06	0.02	−0.15**	−0.06	0.84
Retired	—	−0.21*	−0.09	—	0.13*	−0.04
Age	—	—	0.00	—	—	−0.02
Gender	—	—	0.00	—	—	0.01
Poorer health	—	—	−0.05*	—	—	0.00
Lower income	—	—	−0.04*	—	—	−0.03*
Problem-drinking history
Current problem drinker	—	—	−0.19*	—	—	−0.16**
Former problem drinker	—	—	0.20*	—	—	0.06
Model fit
Deviance (−2LL)	10,989.4**	10,977.3**	10,843.2**	9,228.9**	9,224.7**	9,023.5**
χ2 difference test (df)		12.1 (2)**	146.2 (14)**		4.2 (2)ns	205.4 (14)**

Notes: LL = log likelihood; ^ns = not significant.

^*p ≤ .05;

^**p ≤ .01.

Participants who had active drinking problems at baseline drank more frequently at that time (intercept = 1.7) and experienced steeper decline in their drinking frequency (slope = −0.19). In contrast, former problem drinkers initially drank less frequently than did other sample members (intercept = −1.5) but also declined more slowly in drinking frequency (slope = 0.20). Addition of the covariates to retired status significantly improved the fit of the predictive model of drinking frequency trajectories.

Regarding quantity of alcohol consumed, the unconditional growth model (Model A) indicated that, at baseline, the average sample member was consuming almost 2.5 oz. of ethanol per drinking occasion and that this amount declined moderately (slope = −0.15) over the 10-year interval. Retired status predicted a steeper rate of decline in alcohol consumption (Model B; slope = −0.13), but this effect disappeared once the model also included as predictors participants’ age, gender, health status, financial resources, and current and former problem-drinker status (Model C). Participants in poorer health and those with past drinking problems initially consumed less alcohol than did other sample members (intercepts = −0.76 and −0.49, respectively). Individuals with lower income at baseline declined significantly more steeply in alcohol consumption (slope = −0.03), as did participants who had active drinking problems at baseline (slope = −0.16). Addition of the covariates to the model resulted in significant improvement of its fit. Subsequent analyses showed no statistically significant interactions between retired status and age, gender, health status, or income; interactions between retired status and drinking problems did not add significantly to improved prediction of 10-year drinking trajectories.

Discussion

This study focused on the effect of retired status on older adults’ long-term within-person drinking trajectories and addressed whether this relationship is accounted for or moderated by age, gender, health, income, and current or former problem-drinker status. Consistent with previous cross-sectional studies and short-term follow-ups, individuals in this sample declined moderately over a 10-year interval in frequency and quantity of alcohol consumption. Considered alone, retired status facilitated a somewhat steeper decline in frequency and amount of ethanol consumed. However, retired status had no effect on drinking trajectories once baseline health, income, and current and former problem-drinker status were considered.

Health status, income, and history of problem drinking at late middle age—rather than retired status—appear to steer the direction of subsequent late-life drinking trajectories. Health status at this age may encompass symptoms, disease prognoses, and medication use that contraindicate alcohol consumption and prevent full participation in social activities at which alcohol is served. Having lower income is likely to constrain the purchase of alcoholic beverages (Galea et al., 2007) and may also restrict participation in social activities that involve alcohol.

Perriera and Sloan (2001) demonstrated the importance of problem-drinking history for predicting the course of late-life alcohol consumption. However, they used a single, baseline CAGE assessment to determine drinking problem history and, therefore, could not distinguish between the effects of current drinking problems and more remote drinking problems on late-life drinking patterns. Our findings highlight the importance of drawing this distinction: Having active drinking problems at baseline foreshadowed a steeper decline in frequency and quantity of alcohol consumption. whereas former problem-drinker status somewhat slowed decline in frequency of drinking. Because these aspects of drinking history had such different effects on long-term drinking trajectories, they should be considered separately in future studies predicting the course of late-life alcohol consumption.

This study has several limitations, including narrow generalizability and its insensitivity to short-term or temporary effects of retirement on late-life alcohol consumption. Statistical regression to the mean may have influenced the findings, and limited sample size may have precluded detection of moderating effects. In addition, it should be noted that we focused on effects of stable retired status as distinct from retirement as an event. Notwithstanding these concerns, this study contributes to previous research by demonstrating that health, income, and problem-drinking history outweigh retired status as key predictors of late-middle-age adults’ long-term within-person drinking trajectories. Further, it suggests that older adults’ drinking problem recency should be considered in future studies of retirement and late-life drinking patterns.