Alcohol Use Predicts 10-Year Depressive Symptom Trajectories in the Health and Retirement Study

Abstract

Objective: To determine effects of late-middle-aged adults’ baseline drinking behavior on their subsequent 10-year depressive symptom trajectories. Method: Health and Retirement Study participants (N = 7,939) were assessed on baseline demographic, health, and drinking characteristics, and biennially assessed for the next 10 years on their depressive symptoms. Results: Growth mixture modeling generated four classes of depressive symptom trajectories: Consistently low (72%), consistently elevated (6%), increasing (12%), and decreasing (10%). Baseline abstinence from alcohol, possibly enforced by poorer health and a history of drinking problems, and heavier drinking, “binge” drinking, and having a history of drinking problems, raised risk of membership in the “consistently elevated” class. Abstinence by participants without history of drinking problems—and light, moderate, and heavier drinking—protected against membership in the “increasing” class. Abstinence by participants without history of drinking problems elevated—and moderate drinking reduced—likelihood of membership in the “decreasing” class. Discussion: Late-middle-aged adults’ alcohol use is associated with the subsequent long-term course of their depressive symptoms.

Keywords

alcohol consumption drinking behavior depressive symptoms

Introduction

Among older adults, depressive symptoms are associated with poorer health, diminished quality of life, and earlier mortality (Ruo et al., 2003; Schulz et al., 2000; Wagner & Short, 2014). Thus, it is important to identify factors that put older adults at elevated risk of higher or increasing levels of depressive symptoms in late life. Identifying individuals in late-middle-age who carry such risk can inform the development and targeting of appropriate interventions to prevent deleterious depressive symptom patterns in later life.

Patterns of Late-Life Depressive Symptom Trajectories

Recent research suggests that most older adults experience a relatively benign course of depressive symptoms in later life, comprising initial low levels of depressive symptoms in late-middle- to early-old age, that are sustained over the next 8 to 12 years (Andreescu, Chang, Mulsant, & Ganguli, 2008; Kuchibhatla, Fillenbaum, Hybels, & Blazer, 2012; Kuo, Lin, Chen, Chuang, & Chen, 2011; Liang et al., 2011). However, these studies also indicate that many older adults experience undesirable depressive symptom trajectories that contrast with this benign pattern. Taken as a group, these investigations show that between 2% and 7% of adults assessed early in old age experience elevated levels of depressive symptoms that will be sustained over the next 8 to 12 years. They also show that an additional 4% to 10% of older individuals experience depressive symptom trajectories characterized by initially low levels of depressive symptoms that will increase over the subsequent several years. Another trajectory pattern demonstrated by these studies indicates that 5% to 8% of older adults enjoy an improving course of depressive symptoms during later life.

Predictors of Depressive Symptom Trajectory Class Membership

Consistent with the idea that individuals’ personal and social resources in earlier life influence the subsequent course of their physical and mental health functioning (e.g., Liang et al., 2011), several demographic and health characteristics of adults have been shown to prospectively influence their membership in undesirable 10- to 12-year depressive symptom trajectories. Older age at baseline assessment elevates risk of membership in 10-year depressive symptom trajectory classes characterized by persisting moderate to high levels of depressive symptoms (Kuo et al., 2011; Liang et al., 2011). Compared with men, women are at heightened risk of membership in classes characterized by chronic higher levels of depressive symptoms or by increases in depressive symptoms (Kuchibhatla et al., 2012; Kuo et al., 2011; Liang et al., 2011). There is mixed evidence regarding the influence of racial background on class membership. Whereas Kuchibhatla and colleagues (2012) found that non-White race decreased the likelihood of having initially high and persisting levels of depressive symptoms, Liang and associates (2011) reported that African Americans and Hispanics were more likely to experience this type of depressive symptom pattern and to exhibit increased levels of depressive symptoms.

Having more personal and social resources, as reflected in being married, and having more education, higher income, and better health at baseline assessment has been shown to protect against experiencing undesirable depressive symptom trajectories over the subsequent 10 to 16 years (Andreescu et al., 2008; Kuchibhatla et al., 2012; Kuo et al., 2011; Liang et al., 2011; Lincoln & Takeuchi, 2010). On the other hand, similar factors (i.e., being married, male, White, and having higher income and better health at baseline) have been shown to prevent membership in “improving” depressive symptom trajectory classes (Kuchibhatla et al., 2012; Liang et al., 2011; Lincoln & Takeuchi, 2010). This apparent contradiction has not been explicitly addressed by these authors, but it may occur because individuals who enter later life with more abundant personal and social resources also have fewer depressive symptoms and thus a limit on how much they can improve in mood over the ensuing years.

Little research has focused on prospective relationships between older adults’ baseline health behaviors, such as their exercise, use of tobacco, and drinking behavior, at entry to later life, and their subsequent depressive symptom trajectories. Studies by Kuo and colleagues (2011), of older Taiwanese adults, and Byers and associates (2012), who studied women age 65 and older, suggest that more frequent physical activity and abstinence from tobacco use at baseline assessment protect against subsequent 10- and 20-year patterns of chronic, high, and increasing levels of depressive symptoms. With respect to drinking behavior, the Kuo et al. (2011) study participants who consumed alcohol at baseline assessment had a slightly higher risk than did nondrinkers of persisting mild depressive symptoms for the next 10 years. Compared with nondrinkers, drinkers also had lower risk of persisting high, or increasing, levels of depressive symptoms over this interval. Similarly, Byers et al. found that older women who were “frequent drinkers” (7 or more drinks per week) at baseline assessment had somewhat lower risk of increasing depressive symptoms over the next 20 years. The dichotomous measures of alcohol consumption used in these studies make their results difficult to interpret. More detailed Information about participants’ quantity of alcohol consumption and histories of drinking problems might help explain why use of alcohol at baseline can predict both desirable and undesirable long-term depressive symptom trajectories.

Drinking Behavior as a Predictor of Depressive Symptom Trajectories

The present study extends the approach of previous investigators (Byers et al., 2012; Kuo et al., 2011) to examine in detail the prospective influence of older adults’ use of alcohol in late-middle-age on their subsequent long-term depressive symptom trajectories. The notion that drinking behavior in later-middle life can influence the subsequent course of older adults’ depressive symptoms is based on the view that drinking in moderation may reflect social connectedness and may have physiological benefits but that heavy drinking, although in part representing an attempt to manage stress and depressive symptoms, is ultimately ineffective in doing so. Instead, heavy drinking may directly and indirectly contribute to increased depressive symptoms through its depressant effect on brain functioning and negative consequences for relationships and everyday functioning. Results of previous cross-sectional and short-term follow-up studies are consistent with this picture, indicating that both quantity of alcohol consumed and qualitative indicators of alcohol use (e.g., “binge” drinking, drinking problems) are associated with depressive symptom levels (e.g., Graham, Massak, Demers, & Rehm, 2007; Sullivan, Goulet, & Fiellin, 2011).

With respect to quantity of alcohol consumption, researchers have shown positive linear, as well as J-shaped relationships, between amounts of alcohol consumed and depressive symptoms. In the J-shaped relationship, abstinence/light and heavier drinking predict higher, and moderate drinking predicts lower, levels of depressive symptoms. Presumably, this is because higher amounts of alcohol, and associated health problems, generate or exacerbate depressive symptoms, whereas more moderate drinking through a variety of mechanisms (e.g., salutary effects of moderate alcohol intake; socializing associated with drinking) encourages better mood (Alati et al., 2005; Caldwell et al., 2002; Degenhardt, Hall, & Lynskey, 2001; Rodgers et al., 2000; Skogen, Harvey, Henderson, Stordal, & Mykletun, 2009).

The relationship between abstinence and elevated depressive symptoms in the J-shaped relationship has been explained by the “sick quitter” effect (Shaper, Wannamethee, & Walker, 1988) in which individuals have stopped drinking in response to elevated drinking problems and health difficulties: These, in turn, are associated with higher depressive symptom levels (Alati et al., 2005). Taken together, these findings suggest that individuals who have low and high extremes of alcohol use in late-middle-age may be at risk of experiencing a longer term course of sustained high, or increasing, levels of depressive symptoms. These findings also underscore the importance of considering history of drinking problems, as well amount of alcohol consumed, in interpreting the relationship between abstinence from alcohol in late-middle-age and subsequent depressive symptom trajectories.

Qualitative aspects of alcohol use, such as episodic heavy or “binge” drinking, and drinking in ways that have negative physical, psychological, or social consequences (i.e., “drinking problems”) may also predict a poorer long-term course of depressive symptoms in later life. These drinking variables have been shown to prospectively influence morbidity, mortality, and depressive symptom outcomes in older and mixed-age samples (e.g., Holahan, Schutte, Brennan, Holahan, & Moos, 2014; Rehm, Greenfield, & Rogers, 2001). However, we are unaware of any research examining the prospective relationship of these qualitative indicators of alcohol use in late-middle-age to older adults’ longer term depressive symptom trajectories.

Purpose of this Study

In this study, we analyze data from a 10-year longitudinal Health and Retirement Study (HRS) sample to determine the prospective influence of late-middle-aged adults’ use of alcohol at baseline assessment, at about age 60, on their subsequent membership in 10-year depressive symptom trajectory classes. To do so, we first identify the number and type of depressive symptom trajectory classes that characterize these HRS data. We expect to find classes of depressive symptom trajectory patterns similar in number and type to those identified earlier in HRS and other mixed-age and older longitudinal health survey samples.

Regarding the relation between baseline alcohol use and subsequent depression trajectories, we expect that late-middle-aged adults who abstain from alcohol or engage in heavier drinking will be at elevated risk of membership in undesirable depressive symptom trajectory classes (e.g., consistently high levels of depressive symptoms; increasing depressive symptoms), as will those who engage in episodic heavy (“binge”) drinking or have histories of drinking problems. In contrast, we predict that consuming alcohol at moderate levels at baseline will protect against membership in undesirable depressive symptom trajectory classes and predict membership in a desirable depressive symptom class, characterized by improved depressive symptom levels.

Method

Sample

Our sample was drawn from the HRS which has comprised longitudinal biennial assessment of the health and economic characteristics of adults age 50+ since 1992 (for details, see Hauser & Willis, 2005; Juster & Suzman, 1995; http://hrsonline.isr.umich.edu/index.php). Because of historical HRS project design changes, and cross-wave differences in HRS health item content, HRS measures of alcohol use did not become completely commensurate across assessment waves until 1996. Therefore, we chose 1996 as the baseline assessment point for this study. We selected from the overall 1996 HRS sample individuals aged 55 to 65 (n = 8,635). Next, we obtained these individuals’ 1996, 1998, 2000, 2002, 2004, and 2006 HRS demographic and health data, resulting in a 10-year, 6-wave HRS longitudinal sample.

Follow-up rates in the baseline sample of 8,635 were high (e.g., 96% and 93%, respectively, at the 1998 and 2000 data collections) and closely mirror those reported for the overall HRS cohort initially interviewed in 1992 (HRS, 2011). We removed from this baseline sample of n = 8,635 participants data of 696 participants whose 1996 information was provided by HRS proxy informants (generally a spouse or other family member reporting for participants in a preceding HRS wave). We did this because drinking behavior and depressive symptoms are key variables in the present investigation and HRS does not ask proxy informants to report this subjective information about participants. Thus, our final longitudinal sample size was n = 7,939.

Measures

Baseline demographic characteristics and medical conditions

HRS demographic and health measures are adapted from established U.S. national health surveys, such as the National Health Interview Survey (NHIS) and the National Health and Nutrition Examination Survey (NHANES; for details, see Fisher, Faul, Weir, & Wallace, 2005). Demographic variables included in this study were participants’ baseline (i.e., in 1996) age, racial background (0 = non-White, 1 = White), gender (0 = male, 1 = female), marital status (0 = unmarried, 1 = married), and income in US$10,000 increments (from 1 = up to US$10,000, to 8 ≥ US$70,000). Because older adults’ health has been shown to influence how much they drink and their risk of drinking problems and depressive symptoms (e.g., Moos, Brennan, & Schutte, 2005; Ruo et al., 2003; Schutte, Nichols, Brennan, & Moos, 2003), we included baseline number of medical conditions as a covariate in this study. This variable was assessed by a count of eight medical conditions (high blood pressure, heart problems, diabetes, lung disease, cancer, stroke, arthritis, and back pain); this summary measure has been shown in previous studies to have good validity as an indicator of health status (Brennan & Greenbaum, 2005; Brennan, Holland, Schutte, & Moos, 2012; Brennan, Kagay, Geppert, & Moos, 2000; Joseph, Ganzini, & Atkinson, 1995).

Baseline drinking behavior

Amount of alcohol

Amount of alcohol was assessed at baseline in 1996 and was based on the number of drinks per day consumed by participants on days they drank during the prior 3 months. Individuals who answered “no” to the question, “Do you ever drink alcoholic beverages . . .?” were assigned a “0” for this measure. Information about participants’ number of drinks per day and their weekly frequency of drinking was used to determine their membership in mutually exclusive categories (1 = yes, 0 = no), based on National Institute on Alcohol Abuse and Alcoholism (NIAAA) recommended drinking guidelines (Gunzerath, Faden, Zakhari, & Warren, 2004; NIAAA, 2007), indicating amount of alcohol consumed. The categories included abstinence, consumption of zero drinks per day (with and without a history of drinking problems, as indicated by CAGE items, described below); light drinkers, men who drank two or fewer drinks per day, less often than once a week and women who drank one drink or less per day, less often than once a week; moderate drinkers, men who drank two or fewer drinks per day, once a week or more and women who drank one drink or less, once a week or more; and heavier drinkers, men who consumed more than two drinks per day, once a week or more and women who consumed more than one drink per day, once a week or more.

Episodic heavy drinking and drinking problems

Episodic heavy drinkers were coded “1” to represent men who drank four or more drinks per day, once a week or more, and women who drank three or more drinks per day, once a week or more at baseline assessment in 1996. Drinking problems were assessed only at participants’ initial HRS interviews, which occurred 3 to 4 years prior to baseline for this study. This assessment used the CAGE instrument (Ewing, 1984; Mayfield, McLeod, & Hall, 1974), a valid screening tool for detection of alcohol problems (Buchsbaum, Buchanan, Centor, Schnoll, & Lawton, 1991; Mayfield et al., 1974) that distinguishes well between individuals with and without drinking problems (Chan, Pristach, & Welte, 1994; McIntosh, Leigh, & Baldwin, 1994). CAGE items tap participants’ responses (0 = no; 1 = yes) to four questions: “Have you ever felt that you should cut down on drinking?”; “Have people ever annoyed you by criticizing your drinking?”; “Have you ever felt bad or guilty about drinking?”; “Have you ever taken a drink first thing in the morning (‘eye opener’) to steady your nerves or get rid of a hangover?” Because HRS participants’ initial interviews took place 3 to 4 years before baseline for this study, and because they were asked only whether they had ever had CAGE drinking experiences, it is not possible to discern from the HRS data whether affirmative CAGE responses reflect currently active, recent-past, or remote-past drinking problems. Therefore, from participants’ answers to CAGE items, we created a dichotomous variable (0 = no; 1 = yes) indicating presence of one or more alcohol-related problems, at some point in life, which we labeled history of drinking problems.

Depressive symptoms

We calculated participants’ depressive symptom scores at each of the six waves of HRS information in our longitudinal data set. HRS used a subset of the Center for Epidemiologic Studies Depression scale (CES-D; Radloff, 1977) items, reformatted to capture “yes” and “no” responses (Steffick, 2000), to assess participants’ depressive symptoms in the past 7 days. We used an averaged count (sum of “yes” responses, divided by number of items) of five of the HRS CES-D items (feeling depressed, having restless sleep, inability to “get going,” enjoying life [reverse scored], and feeling happy [reverse scored]) to represent participants’ depressive symptoms at each of their biennial assessments. We used these five items because they are almost identical to five depressive symptom items included in another longitudinal community sample of older adults in which we are examining later-life relationships between alcohol use and health (e.g., Brennan, Schutte, SooHoo, & Moos, 2011) and comparing them with HRS findings. The HRS CES-D has been shown to have good construct validity (Steffick, 2000); average cross-wave Cronbach’s alpha for our five-item depressive symptom measure was .70.

Analytic Approach

We used SPSS 21.0 and Mplus 7.11 (L. K. Muthén & Muthén, 1998-2012) software to analyze the data. We first conducted descriptive statistical analyses to determine the sample’s baseline demographic and drinking characteristics. Next, we conducted growth mixture modeling (GMM) of the depressive symptom data. Conventional growth modeling assumes that longitudinal outcomes under investigation come from a single population, and thus a single-population growth model best accounts for variation in sample members’ longitudinal outcome trajectories. In contrast, GMM assumes that there may be several heterogeneous subpopulations within the overall population and that several distinct subtypes of trajectory are a better fit to the data describing sample members’ longitudinal outcomes than is a single monolithic growth model (B. Muthén et al., 2002; B. Muthén & Muthén, 2000).

Following previous investigators, we conducted GMM to identify the number and type of 10-year depressive symptom trajectory classes characteristic of our sample. GMM Mplus 7.11 software uses full information maximum likelihood estimation to generate growth trajectory parameters; this is a widely accepted statistical approach to handling missing data in longitudinal data analyses (Enders, 2010; Lincoln & Takeuchi, 2010; B. Muthén & Shedden, 1999). Moreover, GMM Mplus modeling routines provides Akaike Information Criterion (AIC) and Adjusted Bayesian Information Criterion (ABIC) model fit criteria, as well as an entropy index, to assist judgments of the correct number of trajectory classes that comprise a set of longitudinal data (Petras & Masyn, 2010).

Following earlier researchers, we began unconditional growth model testing by considering whether linear or quadratic trajectory models best described our sample’s 10-year depressive symptom trajectories; results showed that a linear model of the trajectories fit the data better than did a quadratic model. We then determined the number of classes of 10-year depressive trajectories characteristic of our data with the constraint of fixed zero variance around the intercept and slope in each trajectory class, rendering our approach a latent class growth analysis (LCGA), a subtype of GMM, which assumes that all individual growth trajectories within a class are homogeneous (Jones, Nagin, & Roeder, 2001; Jung & Wickrama, 2008).

Following Petras and Masyn (2010), we used “elbow tests” of diminishing then increasing AIC and ABIC criteria, and of increasing then diminishing entropy values, as well as considerations of the size and substantive meaning of classes, to judge the number of classes that best represent the structure of the 10-year depressive symptom data examined here. Finally, we conducted multinomial logistic regression analyses to estimate the effects of participants’ baseline demographic and drinking characteristics on their likelihood of belonging to the depressive symptom trajectory classes identified with GMM.

Results

Baseline Sample Characteristics

As shown in Table 1, the baseline sample comprised 3,484 men (44%) and 4,455 women (56%); the average age of the sample was 59.80 years (SD = 3.16). The majority of the sample (80%) was White, and most of the participants were married (73%). Participants’ average income was 4.16 (SD = 2.19), indicating that it ranged from US$30,000 to US$40,000. At baseline, participants reported an average of 1.75 (range = 0-8; SD = 1.40) medical conditions.

Table 1.

Descriptive Statistics: Demographic and Drinking Variables in the Overall Sample (N = 7,939).

Variables
Demographic
Age (M, SD)	59.80 (3.16)
Gender
Male	43.9% (3,484)
Female	56.1% (4,455)
Race/ethnicity
White	80.0% (6,355)
Other	20.0% (1,584)
Marital status
Married	73.1% (5,804)
Unmarried	26.9% (2,135)
Income (M, SD)	4.16 (2.19)
Number of medical conditions (M, SD)	1.75 (1.40)
Drinking
Amount of alcohol consumed
Abstinent without history of drinking problems	39.5% (3,133)
Abstinent with history of drinking problems	9.4% (746)
Light drinkers	17.0% (1,352)
Moderate drinkers	20.0% (1,586)
Heavy drinkers	14.1% (1,122)
Episodic heavy (“binge”) drinking (% yes)	5.5% (439)
History of drinking problems (% yes)	23.7% (1,879)

At baseline, almost half of this HRS sample comprised nondrinkers; about 20% of these nondrinkers had a history of drinking problems. Light drinkers comprised 17% of the sample; 20% were moderate drinkers and 14% were heavy drinkers. Almost 6% of participants engaged in episodic heavy drinking and about 24% had a history of drinking problems according to their CAGE responses.

Identification of Depressive Symptom Trajectory Classes

Table 2 shows the change in AIC and ABIC model fit criteria and entropy that occurred with each class addition. Changes in the model fit criteria and in entropy indicate that a four-class model of 10-year depressive symptom trajectories provides the best fit to the data. Class 1 participants (72% of the sample; n = 5,779) exhibited a pattern of consistent, low levels of depressive symptoms over the 10 years of follow-up. In this “consistent low” class, the average trajectory intercept was .07 (p < .01) with slope = .00. Class 2 participants (6% of the sample; n = 456) showed a pattern of initially higher and nondeclining levels of depressive symptoms over the 10-year interval. In this “consistently elevated” class, the average trajectory intercept was .70 (p < .01) and its slope = −.01 (ns). Class 3 participants (12% of participants; n = 930) exhibited initially lower but increasing depressive symptom levels over follow-up; in this “increasing” class, the average trajectory intercept was .21 (p < .01) with slope = .03 (p < .01). Class 4 comprised 10% of the sample (n = 773). In this “decreasing” class, the average trajectory intercept was .49 (p < .01) with slope = −.03 (p < .01).

Table 2.

Model Fit Statistics and Growth Parameters for the Depression Trajectory Classes.

Class	AIC	ABIC	Entropy	n	%	Intercept	Slope
1	2,557	2,613	—	c1: 7,939	100	0.17**	−0.00**
2	−9,083	−9,006	0.91	c1: 6,573	83	0.10**	−0.00
				c2: 1,360	17	0.50**	0.00
3	−11,170	−11,072	0.86	c1: 5,854	73	0.08**	−0.00
				c2: 534	7	0.67**	−0.00
				c3: 1,550	20	0.32**	0.00
4	−12,136	−12,017	0.86	c1: 5,779	72	0.07**	0.00
				c2: 456	6	0.70**	−0.01
				c3: 930	12	0.21**	0.03**
				c4: 773	10	0.49**	−0.03**
5	−12,570	−12,431	0.84	c1: 5,560	70	0.07**	0.00
				c2: 319	4	0.79**	−0.02**
				c3: 904	11	0.16**	0.03**
				c4: 751	10	0.50**	−0.04**
				c5: 405	5	0.38**	0.03**

Note. AIC = Akaike Information Criterion; ABIC = Adjusted Bayesian Information Criterion.

p < .01.

Predictors of Trajectory Class Membership

Baseline demographic characteristics and medical conditions

Results of multinomial logistic regressions (Table 3) show the effects of participants’ baseline demographic characteristics and number of medical conditions on their likelihood of belonging to the “consistently elevated,” “increasing,” and “decreasing” depressive symptom trajectory classes relative to belonging to the reference class of individuals with consistently low levels of depressive symptoms. Non-White participants, women, and those who were unmarried, had lower income, and had more medical conditions at baseline assessment were at higher risk of belonging to the “consistently elevated” depressive symptom trajectory class. For example, being a woman more than tripled the chances of belonging to this class and having more medical conditions almost doubled it.

Table 3.

Multinomial Logistic Regressions: ORs for the Effects of Baseline Demographic Characteristics and Medical Conditions on Depressive Symptom Trajectory Class Membership.

Predictors	OR
Demographic variables
Age
Consistently Elevated versus Consistently Low	0.99
Increasing versus Consistently Low	0.97
Decreasing versus Consistently Low	1.06
White ethnicity
Consistently Elevated versus Consistently Low	0.59**
Increasing versus Consistently Low	0.84
Decreasing versus Consistently Low	0.57**
Female gender
Consistently Elevated versus Consistently Low	3.86**
Increasing versus Consistently Low	1.22
Decreasing versus Consistently Low	0.81
Married status
Consistently Elevated versus Consistently Low	0.34**
Increasing versus Consistently Low	0.71*
Decreasing versus Consistently Low	0.44**
Higher income
Consistently Elevated versus Consistently Low	0.67**
Increasing versus Consistently Low	0.80**
Decreasing versus Consistently Low	0.77**
More medical conditions
Consistently Elevated versus Consistently Low	1.95**
Increasing versus Consistently Low	1.56**
Decreasing versus Consistently Low	1.64**

Note. OR = odds ratio;Consistently Elevated = consistently elevated depressive symptoms trajectory class; Consistently Low = consistently low depressive symptoms trajectory class; Increasing = increasing depressive symptoms trajectory class; Decreasing = decreasing depressive symptoms trajectory class.

p < .01.

Being unmarried, having lower income, and having more medical conditions at baseline also increased the risk of belonging to the “increasing” depressive symptom trajectory class. The likelihood of belonging to the “decreasing” depressive symptom class was enhanced by being non-White, unmarried, lower income, and having more medical conditions.

Baseline drinking behavior

As shown in Table 4, being abstinent at baseline increased the likelihood of being in the “consistently elevated” depressive symptom class. Whereas being abstinent without history of drinking problems doubled the likelihood of belonging to this class (odds ratio [OR] = 2.13), being abstinent with a history of drinking problems increased it almost fivefold (OR = 4.88).

Table 4.

Multinomial Logistic Regressions: ORs for the Effects of Baseline Drinking Behavior on Depressive Symptom Trajectory Class Membership.

Predictors	OR	Adjusted OR
Drinking variables
Abstinence without history of drinking problems
Consistently elevated versus consistently low	2.13**	1.07
Increasing versus consistently low	0.66**	1.30
Decreasing versus consistently low	1.63**	1.13
Abstinence with a history of drinking problems
Consistently elevated versus consistently low	4.88**	1.38
Increasing versus consistently low	1.44	2.48**
Decreasing versus consistently low	1.66	1.90**
Light drinkers
Consistently elevated versus consistently low	1.51	0.80
Increasing versus consistently low	0.34**	0.76
Decreasing versus consistently low	0.66	0.66
Moderate drinkers
Consistently elevated versus consistently low	1.09	0.40**
Increasing versus consistently low	0.46**	0.59**
Decreasing versus consistently low	0.22**	0.65
Heavier drinkers
Consistently elevated versus consistently low	1.76**	1.14
Increasing versus consistently low	0.43**	0.95
Decreasing versus consistently low	1.29	1.50
Episodic heavy drinkers
Consistently elevated versus consistently low	2.71**	1.84**
Increasing versus consistently low	0.75	1.54
Decreasing versus consistently low	1.17	1.31
History of drinking problems
Consistently elevated versus consistently low	3.35**	1.81**
Increasing versus consistently low	1.34	1.45
Decreasing versus consistently low	1.12	1.45

Note. OR = odds ratio; Adjusted OR = odds ratio after statistically adjusting for number of medical conditions; consistently elevated = consistently elevated depressive symptoms trajectory class; consistently low = consistently low depressive symptoms trajectory class; Increasing = increasing depressive symptoms trajectory class; decreasing = decreasing depressive symptoms trajectory class.

p < .01.

Drinking more heavily, heavy episodic or “binge” drinking, and having a history of drinking problems, also elevated the risk of belonging to the “consistently elevated” depressive symptom trajectory class. Being a heavier drinker almost doubled the likelihood of belonging to this class (OR = 1.76) and being an episodic heavy or “binge” drinker raised it almost threefold (OR = 2.71). Drinking in a problematic manner currently or in the past (i.e., affirming one or more of the CAGE items at baseline) more than tripled the chances of belonging to the “consistently elevated” depressive symptom trajectory class.

With respect to membership in the “increasing” depressive symptom trajectory class, being a nondrinker without history of drinking problems was protective against membership in this class (OR = .66), as was light (OR = .34), moderate (OR = .46), and heavier (OR = .43) drinking. Whereas being abstinent without a history of drinking problems increased the likelihood (OR = 1.63) of belonging to the “decreasing” depressive symptom trajectory class, being a moderate drinker reduced it (0.22).

In general, statistically controlling for participants’ number of medical conditions at baseline diminished the influence of baseline drinking behavior on risks of belonging to undesirable depressive symptom trajectory classes (Table 4). For example, when number of medical conditions at baseline was statistically controlled, the risk of belonging to the “consistently elevated” depressive symptom trajectory class attributable to being abstinent with a history of drinking problems was diminished from 4.88 to 1.38 (ns); the predictive effect of being a heavier drinker declined from 1.76 to 1.14 (ns), and that of being an episodic heavy or “binge” drinker declined from 2.71 to 1.14 (ns), respectively.

Discussion

Our findings add to accumulating evidence that most older adults experience the later life course free of serious depressive symptoms. Nevertheless, a significant minority of older adults experience undesirable depressive symptom patterns in later life; membership in these pattern groups is predictable from the personal and social resources, as well as the use of alcohol and history of drinking problems, of late-middle-aged adults entering later life.

Our results closely resemble those of previous investigators who have shown that the course of depressive symptoms in later life is not monolithic but characterized instead by multiple classes of depressive symptom trajectories. For example, a number of previous studies have used LCGA and GMM to show that, although the predominant pattern of depressive symptoms in later life is consistent low levels of depressive symptoms over 8- to 12-year intervals, significant numbers of older adults fall into symptom trajectory classes that are distinct from this pattern (Andreescu et al., 2008; Kuchibhatla et al., 2012; Kuo et al., 2011; Liang et al., 2011).

Not only did our analyses reveal trajectory classes similar to those in earlier studies, they also yielded similar results with respect to the relative size of each of the classes. We found that 6% of our HRS sample had a pattern of initially elevated depressive symptoms that were sustained over the next 10 years. In previous investigations (Andreescu et al., 2008; Kuchibhatla et al., 2012; Kuo et al., 2011; Liang et al., 2011), the percentage in this class ranged from 2% to 7%. Compared with these earlier studies, which reported a range of 4% to 10% of participants in “increasing” depressive symptom trajectory categories, our results showed that 12% of HRS participants experienced an increase over time in depressive symptoms. Similarly, whereas earlier research (Andreescu et al., 2008; Kuchibhatla et al., 2012; Kuo et al., 2011; Liang et al., 2011) indicated improved depressive symptom trajectories among 5% to 8% of their samples, 10% of our HRS sample showed a decrease over time in depressive symptoms.

Our results also generally replicate those of earlier investigations with respect to protective effects of better personal and social resources at late-middle-age on subsequent membership in undesirable depressive symptom trajectory classes. In this regard, our findings are similar to those of several researchers (Andreescu et al., 2008; Kuchibhatla et al., 2012; Kuo et al., 2011; Liang et al., 2011; Lincoln & Takeuchi, 2010) in showing that at late-middle-age, being male, White, married, and having higher income and better health reduced individuals’ risk of experiencing deleterious symptom trajectory patterns (sustained elevated or increasing symptoms) over the next 10 to 16 years. However, we found that these same factors prevented membership in the “decreasing” (i.e., “improving”) depressive symptom trajectory class. Previous investigators have reported similar counterintuitive findings but have not explicitly addressed them (Kuchibhatla et al., 2012; Liang et al., 2011; Lincoln & Takeuchi, 2010). One possible explanation for them is that individuals who enter later life with more abundant personal and social resources, as reflected by higher income, better health, and being married, also tend to enter it with fewer depressive symptoms (Areán & Reynolds, 2005; Blazer & Hybels, 2005), so there is less room for improvement in these individuals’ mood during subsequent years.

In the same vein, individuals who begin later life with more personal and social resources may be less compelled at late-middle-age to initiate behavioral and personal changes to improve their health and mood. For example, late-middle-aged adults with health problems may be forced to initiate health behavior changes, such as reduction or cessation of their alcohol intake, improved eating habits, and increased exercise, to improve their physical functioning. Improved health that results from these behavioral changes may also be associated with an improved long-term course of depressive symptoms.

Our results extend earlier research in this area by highlighting late middle-aged adults’ use of alcohol as a key health behavior that can affect the long-term course of their depressive symptoms. Our findings provided partial support for our hypotheses concerning the prospective relations of late-middle-aged adults’ drinking behavior with their subsequent depressive symptom trajectories. Consistent with our predictions, we found that baseline abstinence from alcohol, heavier drinking, engaging in episodic heavy or “binge” drinking, and having a history of drinking problems significantly increased the risk of membership in a “consistently elevated” 10-year pattern of depressive symptoms. This is consonant with results of earlier investigations showing J- or U-shaped relationships between amounts of alcohol consumed and poorer health outcomes. For example, abstinence from alcohol and heavier drinking have both been shown to elevate risk of morbidity and mortality (Alati et al., 2005; Caldwell et al., 2002; Degenhardt et al., 2001; Rodgers et al., 2000; Skogen et al., 2009).

The U-shaped risk of membership in the “consistently elevated” depressive symptom trajectory class—that is, the much elevated risk of membership in this class among both baseline nondrinkers and baseline heavier drinkers—may be accounted for in part by a “sick quitter” effect (Shaper et al., 1988). Some late-middle-aged individuals may stop drinking due to health-related and personal difficulties wrought by their histories of drinking problems. A legacy of ongoing health and interpersonal difficulties related to earlier drinking problems may continue to elevate risk of sustained, higher levels of depressive symptoms despite these individuals’ abstinence from alcohol. This possibility is supported by the findings that risk of membership in the “consistently elevated” depressive symptom trajectory class was lower among nondrinkers with no history of drinking problems (OR = 2.13) than among nondrinkers with a history of drinking problems (OR = 4.88), and that these odds ratios were significantly diminished by statistically controlling for participants’ baseline number of medical conditions.

Our findings also underscore the importance of examining qualitative as well as quantitative aspects of drinking behavior in seeking to understand the connections between alcohol use in late-middle-age and the subsequent course of late-life depressive symptoms. Qualitative ways of using alcohol, such as engaging in episodic heavy or “binge” drinking, and drinking in ways that illicit negative physical, psychological, or social consequences (“drinking problems”) foreshadowed membership in the consistently elevated depressive symptom trajectory class, and this relationship remained significant even when number of baseline medical conditions were considered. Qualitative aspects of drinking behavior appear to be at least as important as baseline amounts of alcohol consumed for predicting the subsequent late-life course of depressive symptoms. Furthermore, our results suggest that considering both quantitative and qualitative elements simultaneously could be productive. Having knowledge of alcohol consumption quantity (abstinence) and drinking problem history improved the ability to predict membership in a trajectory class indicating an increasing pattern of depressive symptoms, even in the context of considering number of health conditions.

With respect to membership in the “increasing” depressive symptom trajectory class, the findings provide partial support for our hypotheses. As predicted, moderate drinking at baseline protected against membership in this class. This finding suggests that moderate drinking has positive mental health outcomes for late-middle-aged adults who adhere to suggested drinking guidelines (e.g., NIAAA, 2007). However, baseline abstinence from alcohol without history of drinking problems, light drinking, and heavier drinking also protected against membership in the “increasing” depressive symptom trajectory class, suggesting that factors in addition to amount of alcohol consumed per se, such as purposeful reduction or cessation of alcohol use to promote health and frequent socializing in leisure contexts that involve alcohol consumption (e.g., family gatherings, dining with friends, traveling), may help explain the protective effects of these amounts of alcohol against membership in the “increasing” depressive symptom trajectory class. In addition, all of these protective effects, except that attributable to moderate drinking, were diminished by statistically controlling for baseline number of medical conditions. This again illustrates ties that bind together physical health, alcohol use, and depressive symptoms in later life. It is possible that improved health associated with being abstinent or drinking lightly at baseline, or being a baseline heavier drinker who quits drinking post-baseline, is the driving force behind the apparent protective effects of abstinence, light drinking, and heavier drinking in late-middle-age against membership in the “increasing” depressive symptoms trajectory class.

Reinforcing the idea that abstinence from alcohol in late-middle-age may have salutary effects on late-life depressive symptom trajectories, we found that abstinence from alcohol among individuals who did not have a history of drinking problems increased the likelihood of membership in the “decreasing” depressive symptom trajectory class. However, we also found that baseline moderate drinking reduced the likelihood of belonging to this class. As suggested earlier, this seeming contradiction may be accounted for by positive lifestyle changes initiated by baseline nondrinkers that result in an improving long-term course of depressive symptoms and limits on improvement in depressive symptoms that can be expected among individuals who are baseline moderate drinkers.

Taken together, our findings highlight the importance of identifying late-middle-aged adults who engage in heavier or binge drinking, or who have histories of drinking problems, and targeting them for intervention efforts to prevent their subsequently experiencing sustained and elevated depressive symptoms. Health care providers and adults themselves in early-old age should be aware that heavier and problem drinking during late-middle-age can foreshadow this depressive symptom pattern and take steps to prevent it. On an optimistic note, our findings suggest that many late-middle-aged adults are engaging in guideline-concordant drinking behavior and drinking cessation efforts that may protect them from increasing depressive symptoms and even lead them to experience fewer depressive symptoms as they age. Thus, it may be beneficial to implement formal treatment and self-help interventions in continued support of these late-life drinking practices. Future research should examine midlife health and personal and social resources as mediators and moderators of the relationship between drinking behavior in late-middle-age and subsequent depressive symptom trajectories; this may help identify which older adults can benefit most from formal treatment and self-help approaches to prevent increasing or sustained depressive symptoms in later life. It also may reveal midlife health and contextual factors that are barriers to, or facilitators of, the success of these interventions.

This study benefits from a large sample size, a multiwave prospective design, use of valid and reliable measures, and application of longitudinal GMM to the HRS depressive symptom data. However, it has several limitations. Although there is good evidence for the reliability and validity of alcohol use and depressive symptom self-reports (e.g., Harris, Wilsnack, & Klassen, 1994; Liu et al., 1996; McHorney, Ware, Rogers, Raczek, & Lu, 1992), the self-reports of alcohol use and depressive symptoms in this study may have been affected by recall bias and other threats to validity (Stone, Shiffman, Atienza, & Nebeling, 2007). Another limitation is that, due to small sample sizes, we could not examine here the effects of specific racial backgrounds (e.g., Latino, Asian, and American Indian background) on membership in undesirable late-life depressive symptom trajectory classes.

Temporal design characteristics of this study may limit the relevance of our findings for design of interventions to prevent ill effects of later-life drinking behavior on depressive symptom patterns. Specifically, this study’s wide time intervals between assessments, and its emphasis on the prospective effects of baseline alcohol use on long-term (10-year) depressive symptom trajectories, may have obscured the existence of important shorter-term relationships between late-life alcohol use and late-life depressive symptoms. Future studies covering shorter time intervals (e.g., weekly or daily), and those that incorporate time-varying measures of alcohol use, health, and depressive symptoms, are needed to better understand relationships between older adults’ use of alcohol and fluctuations in their depressive symptom levels.

A further consideration is that our sample excluded data from HRS proxy informants. Because individuals who required proxy informants were somewhat more likely to be male, unmarried, non-White, and to have fewer medical conditions than were other sample members, our findings cannot be generalized to describe all U.S. adults who were age 60 and older and aging over the interval 1996-2006. Finally, although our prospective design has the strength of temporal precedence in its favor, we are limited in the conclusions we can draw about the causal relationship between late-middle-aged adults’ use of alcohol and the subsequent course of their depressive symptoms. For example, several of our findings imply that factors associated with alcohol use in late-middle-age, such as health problems, levels of socializing, and self-help health behaviors, may be as important as amount of alcohol consumed per se, as determinants of late-life depressive symptom trajectories.

In spite of its limitations, this study contributes to the literature on health and aging by adding to a growing body of work showing that multiple classes characterize the course of depressive symptoms in later life and that drinking-related behavior can help predict membership in these classes. It holds promise for generating future clinically relevant information for determining how individuals’ adult life span drinking behaviors can be changed to improve the course of their late-life depressive symptom patterns.

Footnotes

Authors’ Note

The views expressed here are those of the authors and do not represent those of the Department of Veterans Affairs or the United States Government.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by National Institutes of Health, National Institute on Alcohol Abuse and Alcoholism Grant R01 AA17477 and by Health Services Research and Development, Department of Veterans Affairs.

References

Alati

Lawlor

D. A.

Najman

J. M.

Williams

G. M.

Bor

O’Callaghan

(2005). Is there really a “J-shaped” curve in the association between alcohol consumption and symptoms of depression and anxiety? Findings from the Mater-University Study of Pregnancy and its outcomes. Addiction, 100, 643-651.

Andreescu

Chang

C.-C. H.

Mulsant

B. H.

Ganguli

(2008). Twelve-year depressive symptom trajectories and their predictors in a community sample of older adults. International Psychogeriatrics, 20, 221-236.

Areán

P. A.

Reynolds

C. F.

(2005). The impact of psychosocial factors on late-life depression. Journal of Biological Psychiatry, 58, 277-282.

Blazer

D. G.

Hybels

C. F.

(2005). Origins of depression in later life. Psychological Medicine, 35, 1-12.

Brennan

P. L.

Greenbaum

M. A.

(2005). Functioning problem behavior and health services use among nursing home residents with alcohol-use disorders: Nationwide data from the VA Minimum Data Set. Journal of Studies on Alcohol, 66, 395-400.

Brennan

P. L.

Holland

J. M.

Schutte

K. K.

Moos

R. H.

(2012). Coping trajectories in later life: A 20-year predictive study. Aging & Mental Health, 16, 305-316.

Brennan

P. L.

Kagay

C. R.

Geppert

J. J.

Moos

R. H.

(2000). Elderly Medicare inpatients with substance use disorders: Characteristics and predictors of hospital readmissions over a 4-year interval. Journal of Studies on Alcohol, 61, 891-895.

Brennan

P. L.

Schutte

K. K.

SooHoo

Moos

R. H.

(2011). Painful medical conditions and alcohol use in later life. Pain Medicine, 12, 1049-1059.

Buchsbaum

D. G.

Buchanan

R. G.

Centor

R. M.

Schnoll

S. H.

Lawton

M. J.

(1991). Screening for alcohol abuse using CAGE scores and likelihood ratios. Annals of Internal Medicine, 15, 774-777.

10.

Byers

A. L.

Vittinghoff

Lui

Hoang

Blazer

D. G.

Covinsky

K. E.

. . . Yaffe

(2012). 20-year depressive symptom trajectories among older women. Archives of General Psychiatry, 69, 1073-1079.

11.

Caldwell

T. M.

Rodgers

Jorm

A. F.

Christensen

Jacomb

P. A.

Korten

A. E.

Lynskey

M. T.

(2002). Patterns of association between alcohol consumption and symptoms of depression and anxiety in young adults. Addiction, 97, 583-594.

12.

Chan

A. W. K.

Pristach

E. A.

Welte

(1994). Detection by the CAGE of alcoholism or heavy drinking in primary care and the general population. Journal of Substance Abuse, 6, 123-135.

13.

Degenhardt

Hall

Lynskey

(2001). Alcohol, cannabis and tobacco use among Australians: A comparison of their associations with other drug use and use disorders, affective and anxiety disorders, and psychosis. Addiction, 96, 1603-1614.

14.

Enders

C. K.

(2010). Applied missing data analysis. New York, NY: Guilford Press.

15.

Ewing

J. A.

(1984). Detecting alcoholism: The CAGE questionnaire. Journal of the American Medical Association, 252, 1905-1907.

16.

Fisher

G. G.

Faul

J. D.

Weir

D. R.

Wallace

R. B.

(2005). Documentation of chronic disease measures in the Health and Retirement Study (HRS/AHEAD). Retrieved from http://hrsonline.isr.umich.edu/sitedocs/userg/dr-009.pdf

17.

Graham

Massak

Demers

Rehm

(2007). Does the association between alcohol consumption and depression depend on how they are measured? Alcoholism: Clinical & Experimental Research, 31, 78-88.

18.

Gunzerath

Faden

Zakhari

Warren

(2004). National Institute on Alcohol Abuse and Alcoholism report on moderate drinking. Alcoholism: Clinical & Experimental Research, 28, 829-847.

19.

Harris

T. R.

Wilsnack

R. W.

Klassen

A. D.

(1994). Reliability of retrospective self-reports of alcohol consumption among women: Data from a U.S. national sample. Journal of Studies on Alcohol, 55, 309-314.

20.

Hauser

R. M.

Willis

R. J.

(2005). Survey design and methodology in the Health and Retirement Study and the Wisconsin Longitudinal Study. In Waite

L. J.

(Ed.), Aging, health and public policy: Demographic and economic perspectives (pp. 209-235). New York, NY: Population Council.

21.

Health and Retirement Study. (2011). Sample sizes and response rates. Retrieved from http://hrsonline.isr.umich.edu/sitedocs/sampleresponse.pdf

22.

Holahan

C. J.

Schutte

Brennan

P. L.

Holahan

Moos

R. H.

(2014). Episodic heavy drinking and 20-year total mortality among late-life moderate drinkers. Alcoholism: Clinical & Experimental Research, 38, 1432-1438.

23.

Jones

B. L.

Nagin

D. S.

Roeder

(2001). A SAS procedure based on mixture models for estimating developmental trajectories. Sociological Methods & Research, 29, 374-393.

24.

Joseph

C. L.

Ganzini

Atkinson

R. M.

(1995). Screening for alcohol use disorders in the nursing home. Journal of the American Geriatrics Society, 43, 368-373.

25.

Jung

Wickrama

K. A. S.

(2008). An introduction to latent class growth analysis and growth mixture modeling. Social & Personality Psychology Compass, 2, 302-317.

26.

Juster

F. T.

Suzman

(1995). The Health and Retirement Study: Data quality and early results [Special Issue]. The Journal of Human Resources, 30, S7-S56.

27.

Kuchibhatla

M. N.

Fillenbaum

G. G.

Hybels

C. F.

Blazer

D. G.

(2012). Trajectory classes of depressive symptoms in a community sample of older adults. Acta Psychiatrica Scandinavica, 125, 492-501.

28.

Kuo

S.-Y.

Lin

K.-M.

Chen

C.-Y.

Chuang

Y.-L.

Chen

W. J.

(2011). Depression trajectories and obesity among the elderly in Taiwan. Psychological Medicine, 41, 1665-1676.

29.

Liang

Xiao

Quiñones

A. R.

Bennett

J. M.

(2011). Multiple trajectories of depressive symptoms in middle and late life: Racial/ethnic variations. Psychology and Aging, 26, 761-777.

30.

Lincoln

K. D.

Takeuchi

D. T.

(2010). Variation in the trajectories of depressive symptoms: Results from the American’s Changing Lives Study. Biodemography and Social Biology, 56, 24-41.

31.

Liu

Serdula

M. K.

Byers

Williamson

D. F.

Mokdad

A. H.

Flanders

W. D.

(1996). Reliability of alcohol intake as recalled from 10 years in the past. American Journal of Epidemiology, 143, 177-186.

32.

Mayfield

McLeod

Hall

(1974). The CAGE questionnaire: Validation of a new alcoholism instrument. American Journal of Psychiatry, 131, 1121-1123.

33.

McHorney

C. A.

Ware

J. E.

Jr. Rogers

Raczek

A. E.

J. F. R.

(1992). The validity and relative precision of MOS short- and long-form health status scales and Dartmouth COOP charts. Results from the Medical Outcomes Study. Medical Care, 30, MS253-MS263.

34.

McIntosh

M. C.

Leigh

Baldwin

(1994). Screening for hazardous drinking: Using the CAGE and measures of alcohol consumption in family practice. Canadian Family Practice, 40, 1546-1553.

35.

Moos

R. H.

Brennan

P. L.

Schutte

K. K.

(2005). Older adults’ health and changes in late-life drinking patterns. Aging & Mental Health, 9, 1-11.

36.

Muthén

Brown

C. H.

Masyn

Khoo

S.-T.

Yang

C. C.

. . . Liao

(2002). General growth mixture modeling for randomized preventive interventions. Biostatistics, 3, 459-475.

37.

Muthén

L. K.

(2000). Integrating person-centered and variable-centered analyses: Growth mixture modeling with latent trajectory classes. Alcoholism: Clinical & Experimental Research, 24, 882-891.

38.

Muthén

Shedden

(1999). Finite mixture modeling with mixture outcomes using the EM algorithm. Biometrics, 55, 463-469.

39.

Muthén

L. K.

Muthén

B. O.

(1998-2012). Mplus user’s guide (7th ed.). Los Angeles, CA: Author.

40.

National Institute on Alcohol Abuse and Alcoholism. (2007). Helping patients who drink too much: A clinician’s guide (Updated 2005th ed.). Retrieved from http://pubs.niaaa.nih.gov/publications/Practitioner/CliniciansGuide2005/guide.pdf

41.

Petras

Masyn

(2010). General growth mixture analysis with antecedents and consequences of change. In Piquero

Weisbud

(Eds.), Handbook of quantitative criminology (pp. 69-100). New York, NY: Springer.

42.

Radloff

L. S.

(1977). The CES-D scale: A self-report depression scale for research in the general population. Applied Psychological Measurement, 1, 385-401.

43.

Rehm

Greenfield

T. K.

Rogers

J. D.

(2001). Average volume of alcohol consumption, patterns of drinking, and all-cause mortality: Results from the US National Alcohol Survey. American Journal of Epidemiology, 153, 64-71.

44.

Rodgers

Korten

A. E.

Jorm

A. F.

Jacomb

P. A.

Christensen

Henderson

A. S.

(2000). Non-linear relationships in associations of depression and anxiety with alcohol use. Psychological Medicine, 30, 421-432.

45.

Ruo

Rumsfeld

J. S.

Hlatky

M. A.

Liu

Browner

W. S.

Whooley

M. A.

(2003). Depressive symptoms and health-related quality of life: The Heart and Soul Study. The Journal of the American Medical Association, 290, 215-221.

46.

Schulz

Beach

S. R.

Ives

D. G.

Martire

L. M.

Ariyo

A. A.

Kop

W. J.

(2000). Association between depression and mortality in older adults. Archives of Internal Medicine, 160, 1761-1766.

47.

Schutte

K. K.

Nichols

Brennan

P. L.

Moos

R. H.

(2003). A ten-year follow-up of former problem drinkers: Risk of relapse and implications of successfully sustained remission. Journal of Studies on Alcohol, 64, 367-374.

48.

Shaper

A. G.

Wannamethee

Walker

(1988). Alcohol and mortality in British men: Examining the U-shaped curve. The Lancet, 2, 1267-1273.

49.

Skogen

J. C.

Harvey

S. B.

Henderson

Stordal

Mykletun

(2009). Anxiety and depression among abstainers and low-level alcohol consumers. The Nord-TrØndelag. Addiction, 104, 1519-1529.

50.

Steffick

D. E.

(2000). Documentation of affective functioning measures in the Health and Retirement Study. Retrieved from http://hrsonline.isr.umich.edu/sitedocs/userg/dr-005.pdf

51.

Stone

A. A.

Shiffman

Atienza

A. A.

Nebeling

(2007). The science of real-time data capture: Self-reports in health research. Oxford, UK: Oxford University Press.

52.

Sullivan

L. E.

Goulet

J. L.

Fiellin

D. A.

(2011). Alcohol consumption and depressive symptoms over time: A longitudinal study of patients with and without HIV infection. Drug and Alcohol Dependence, 1, 158-163.

53.

Wagner

D. C.

Short

J. L.

(2014). Longitudinal predictors of self-rated health and mortality in older adults. Preventing Chronic Diseases, 11, Article 130241.