Classes of Individual Growth Trajectories of Religious Coping in Older Adulthood

Abstract

For many individuals, religion provides important cognitive resources for coping with stressors, especially in older adulthood. Although older adults are thought to make more use of these coping strategies than those at younger ages, less is known about how patterns of use change during the span of older adulthood. In a largely Christian sample of U.S. older adults, positive and negative religious coping were measured between 2 and 5 times over a period of 11 years (N = 1,075). Growth mixture modeling extracted latent classes of growth. The optimal solution for positive coping indicated a five-class structure (high, stable; high, declining moderately; high, declining rapidly; low, increasing; and low, stable) and the optimal negative coping solution had three classes (low, declining; low, increasing; and high, declining). Nominal logistic regression examined the relationship of individual characteristics with latent class. Education, religious commitment, religious attendance, and religious doubt were related to positive coping trajectory class. Only religious doubt was related to negative coping class.

Keywords

religion religious coping aged heterogeneity growth mixture modeling

Religious belief provides cognitive resources for coping with stressors for many people (Harrison, Koenig, Hays, Eme-Akwari, & Pargament, 2001). A large majority of the adult U.S. population reports using various religious coping techniques in response to stressors including traumatic experiences (Park & Cohen, 1993; Schuster et al., 2001), illness (Harrison et al., 2001; Koenig, Pargament, & Nielsen, 1998; Tepper, Rogers, Coleman, & Malony, 2001), and stressful life events (Pargament, 1997). Older adulthood is a life stage in which a number of serious stressors, such as worsening health, the death of loved ones, and financial strain, often become more prevalent, while reactivity to stress is also prone to increase (Hay & Diehl, 2010; Mroczek & Almeida, 2004). Combined with the observation that interest in and practice of religion tends to be higher among older adults (Krause, 2008), it is therefore not surprising that several studies have found that older adults tend to rely more on religious forms of coping, in comparison to younger and middle adults (Courtenay, Poon, Martin, Clayton, & Johnson, 1992; Levin, Taylor, & Chatters, 1994). However, some key questions regarding religious coping in late life remain unanswered. First, little is known about how the use of religious coping changes within the span of older adulthood itself. As life expectancies continue to increase, the population of the oldest-old is expected to expand faster than any other segment of the population in the coming decades (U.S. Census Bureau, 2012). As such, many older people can expect to live for another 20 or 30 years after entering older adulthood, and it is likely that they will continue to undergo change and development throughout this span. Second, beyond age and race (Chatters, Taylor, Jackson, & Lincoln, 2008), little is known about factors that may contribute to differences in the use of religious coping among older adults, let alone trajectories of change. In this study, we take advantage of longitudinal data covering older adults, as they move through this life stage to examine different patterns of individual change in religious coping during this period and their potential causes.

Religious Coping

Religious coping is extremely prevalent among the U.S. population as a whole, although it is especially common among subgroups including African Americans (Chatters et al., 2008) and Evangelical Christians (Alferi, Culver, Carver, Arena, & Antoni, 1999). For example, the proportion of those saying that prayer is a “very important” method of dealing with stress has been estimated at 67% of non-Hispanic Whites and 90% of African Americans (Chatters et al., 2008). A large body of research has been devoted to the variety of specific elements of religious belief that can serve to remediate the experience of stress (Harrison et al., 2001). These can be broadly categorized as positive strategies (i.e., those that seek to reduce stress by dealing effectively with the stressor) and negative strategies (i.e., those that tend to exacerbate the stressor; Ano & Vasconcelles, 2005). Pargament, Koenig, and Perez (2000) articulated a taxonomy of 21 of these specific religious coping methods. For example, one coping method would be to cognitively reconstruct the stressful situation in a positive light, in a way that helps create and reinforce a sense of religious meaning (e.g., seeing it as a divine test of one’s faith and thus evidence that God is involved in one’s life). Conversely, cognitive reconstrual can take on a negative form when it creates additional stressors such as seeing the event as a form of punishment by God.

Late Life Changes

A number of studies have found that religious coping is more common among older adults than among those in younger age-groups (Courtenay et al., 1992; Levin et al., 1994). Although a likely explanation for this finding would seem to be that people begin to use religious coping more frequently as they get older, the cross-sectional nature of much of the data in this area makes this interpretation uncertain. Other studies suggest that older birth cohorts have been enduringly more religiously involved in general than their counterparts born more recently (Pew Forum on Religion and Public Life, 2012), and hence it is possible that higher observed levels of religious coping in older groups is the result of these cohort effects. However, a handful of longitudinal studies help to substantiate the idea that religious involvement also increases with age within individuals, especially between middle and older adulthood (Hayward & Krause, 2013; Schwadel, 2011), lending some credence to the idea that there may be within-individual changes in the use of religious coping as well.

This pattern would make a great deal of sense from a life course developmental perspective. According to this general viewpoint, there are changes that occur in older adulthood as a result of factors including changing social roles (e.g., retirement at work and fewer active parental responsibilities) and the growing salience of one’s own mortality that tend to lead to changes in motivations, cognitions, and other key psychological elements. For example, the Eriksonian theory of adult development holds that the key challenge of late life is that of ego integrity versus despair (Erikson, 1968). Religious belief, and especially positive religious coping practices, may be an important way of creating and enacting meaning, thus helping to avoid despair (Hoare, 2002). Similarly, the gerotranscendence perspective (Tornstam, 1997) holds that, during the course of older adulthood, people tend to disengage from material concerns in favor of “transcendent” meaning-related cognition, of which religious coping may be an integral part. Likewise, as mortality becomes more salient in late life, due to deteriorating health, an increasing number of deaths among one’s acquaintances, and a general awareness of one’s own advancing age, the meaning-making functions of religion may become more salient as a means of staving off the risk of death anxiety (Vail et al., 2010).

Although these perspectives are generally consistent in suggesting that the use of religious coping may be likely to continue to increase across the span of older adulthood, the possibility that there are important individual differences in the nature of these changes cannot be ignored. The observation that the dispersion of traits in the population is often higher among older adults than in younger samples is known as the aged heterogeneity perspective (Nelson & Dannefer, 1992). Individuals follow distinct trajectories of change and development throughout the life course and thus it follows that in many cases, their differences will tend to become more pronounced over time as they continue to develop in divergent directions. Equally, diversity in the social environments and in the types of stressors faced by different individuals in older adulthood may promote divergence in the use of religious coping. For example, while highly religious individuals may prefer religious coping practices and thus increase their use in response to the challenges of aging, those who are only marginally religious may choose to focus on other nonreligious strategies when it comes to coping with the same situations, such as turning to family members and friends for social support, and thus may decrease their use of religious coping.

Understanding these patterns and changes is potentially important because the use of religious coping techniques has been found to be closely linked with mental and physical health outcomes, particularly among older adults (Ano & Vasconcelles, 2005; Pargament, Koenig, Tarakeshwar, & Hahn, 2004). Positive coping may have protective effects because it serves to buffer stress. The experience of stress is not only harmful for mental health but also has direct physiological effects operating via hormonal and immunological pathways (McEwen, 2008), exacerbating common physical problems of late life, such as cardiovascular disease, and suppressing resistance to infection. Conversely, negative religious coping responses may have just the opposite effects, increasing stress levels and thereby exacerbating risk for mental and physical problems. Certain forms of negative religious coping may also have deleterious effects because they discourage constructive coping mechanisms. For example, passive religious deferral refers to the tendency to cope with stress by assuring oneself that God is in control of the situation. While this may serve to reduce stressful rumination, it may also lead to negative repercussions in cases where positive effort might be effective in dealing with the source of the stress, as when one avoids seeking potentially beneficial medical treatment for a disease because of the belief that God will solve the problem (Pargament, Koenig, & Perez, 2000). Understanding how the tendency to rely upon both positive and negative religious coping strategies may change over the course of late life, as well as its causes, may thus be important in understanding how certain health risks change in different subpopulations during this phase of life.

Hypotheses

The two primary purposes of this study are to map differences in individual trajectories of change in religious coping practices across the course of older adulthood (aged 65+ years) and to examine the individual characteristics that may be predictive of the course of this trajectory. Consistent with the adult developmental perspective (Erikson, 1968; Tornstam, 1997), we anticipate that the average older adult will exhibit an increase in positive religious coping (Hypothesis 1a) and a decrease in negative religious coping (Hypothesis 1b) during the same time frame. However, consistent with the aged heterogeneity perspective (Nelson & Dannefer, 1992), we also expect that there will be significant intraindividual variation in trajectories, with others following patterns of decline and stability in positive (Hypothesis 2a) and negative (Hypothesis 2b) religious coping. Finally, we propose three exploratory analyses examining the extent to which these between-person differences in change may be associated with factors that include demographics (Hypothesis 3), experience of key stressors (Hypothesis 4), and personal religious background (Hypothesis 5).

Method

This study is based on data from a longitudinal survey regarding religion and health outcomes in older adulthood (Krause, 2002). The first wave was conducted in 2001 and was designed to be representative of the U.S. population aged 66 and older (N = 1,500). The initial sample was drawn from the beneficiary list maintained by the Centers for Medicare and Medicaid Services (see Krause, 2002 for details regarding sampling procedures). The sample was filtered in two ways. First, it was limited to individuals identifying their race as either White or Black. Black individuals oversampled to allow for power to make comparisons between these groups. Second, it was limited to individuals whose religious backgrounds met one of the following criteria: (1) currently identifying as Christian, (2) formerly identifying as Christian, but currently not religious, or (3) never identified with a religious group. This limitation was imposed because a number of religious measures were developed and validated for use only in Christian samples. Subsequent waves were conducted in 2004 (N = 1,024), 2006 (N = 969), and 2008 (N = 718), with a fifth wave in 2013 that included both longitudinal recontact of the surviving initial sample (N = 229) and new cohort refreshing the sample who were contacted for the first time (N = 1,318). Because this study focuses on trajectories of change, the Wave 5 refresher sample is excluded from these analyses. All interviews were conducted face-to-face in participants’ homes. To be included in the analyses, a participant must have been included in at least two waves of data collection, leaving the final sample size of 1,084.

Measurement

Religious coping

A 10-item scale (Krause, 2002; Pargament et al., 2000) was used to measure positive and negative religious coping. Each item was scored on a 4-point response scale from “not at all” to “a great deal,” regarding how often they used specific coping strategies when dealing with major problems in life. The positive religious coping subscale consisted of 5 items (e.g., “I looked to God for strength”), M = 17.0, SD = 3.5, and Cronbach’s α = .86. The negative religious coping subscale consisted of the remaining 5 items (e.g., “I thought it was God’s way of punishing me for things I have done wrong”), M = 7.3, SD = 3.1, and Cronbach’s α = .73. To track different potential trajectories of change in coping, data for these variables were taken from all available waves. Because the focus of this study is on prediction individual trajectories, all other study variables are taken from the first wave only.

Health and well-being

Subjective health was measured using a 3-item scale in which participants rated their health (1) overall, (2) in comparison to others of their age, and (3) in comparison to their own health a year previously. The first of these items was scored on a 4-point response scale from “poor” to “excellent,” whereas the last two were each scored on a 3-point response scale (worse, about the same, or better), M = 7.1, SD = 1.6, and Cronbach’s α = .61.

Symptoms of depression were measured using a 8-item version of the Center for Epidemiologic Studies–Depression (Radloff, 1977), which is used to screen for depression in general samples. Each item is scored on a scale from 0 to 3, with higher scores indicating more severe psychological distress, M = 4.4, SD = 4.9, and Cronbach’s α = .89.

Religious factors

Religious denomination was coded according to the system proposed by Steensland et al. (2000), which divides group affiliations into Catholic, Mainline Protestant, Evangelical Protestant, and other categories. Participants reported their specific denominational affiliations, which were then recoded into these broad categories.

Attendance at religious services was measured using a single item. Participants reported their frequency of attendance on a 9-point scale from never to more than once a week, M = 5.7 and SD = 2.7.

Extent of religious commitment was measured using a 3-item scale (Fetzer Institute & National Institute on Aging, 1999), regarding the importance of religion in one’s life (e.g., “my faith shapes how I think and act each and every day”). Reponses were on a 4-point scale from strongly disagree to strongly agree, M = 10.0, SD = 3.4, and Cronbach’s α = .92.

Religious doubt was measured using a 5-item scale (e.g., “how often do you have doubts about your religious or spiritual beliefs”). This scale was created with the extensive item development procedures discussed by Krause (2002). Responses were on a 4-point scale from never to very often, M = 4.3, SD = 1.8, and Cronbach’s α = .75.

Demographics

Demographic factors measured in this study included race (see the previous-mentioned sample description), age, gender, and years of educational attainment (M = 11.3 and SD = 3.5).

Analytic Approach

The primary analyses in this study included two steps. First, growth mixture modeling (GMM) was used to extract categories of individual change trajectories for use of positive and negative religious coping over the course of older adulthood. The GMM procedure (Jung & Wickrama, 2008; Muthén & Muthén, 2000) is a method in which a latent class model is applied to a linear growth curve, such that the latent category indicates a given number of distinct combinations of intercept and slope of change over time (i.e., different within-person patterns of change). It differs from a latent class growth analysis in that individual trajectories are allowed to vary randomly around the class mean, rather than being assumed to be fixed. The GMM analyses produce estimates of the mean intercept and linear slope effects within each latent class. Additionally, based on the same model, they predict class membership for each individual participant.

The predicted class membership values produced by the GMM analyses were used as the key dependent variable in the second set of analyses. In this step, nominal logistic regression was used to examine the association of key baseline (Wave 1) variables with subsequent trajectory of change in positive and negative religious coping (latent class membership).

Missing Data Imputation

Multiple imputation of missing data was conducted using the Markov Chain Monte Carlo (MCMC) method to deal with missing data because 22.9% of the sample would have been lost in the final nominal logistic regression analyses if listwise deletion of cases with any values missing due to item nonresponse was used. Using MCMC, each value is imputed by simulating a series of random draws from the observed distribution of the target variable, based on the joint distribution of all other variables in the model (Horton & Lipsitz, 2001). All analyses are then conducted separately using each set of imputed values, and the results are pooled. To ensure that the GMM analyses included only direct observations, religious coping values were not imputed, and the values for the independent model variables were imputed at Wave 1 only. To maximize relative efficiency of the analyses, 25 sets of imputations were produced (Graham, Olchowski, & Gilreath, 2007).

Results

GMM results were obtained using data on age and religious coping from all five survey waves. To be included in the analyses, an individual had to have religious coping scores in at least two waves. After excluding individuals with coping responses in only a single wave, the GMM sample contained 3,901 observations from 1,084 individuals. The 416 individuals lost to attrition after the only one wave differed from the remaining sample in terms of age, t(1,484) = 6.98, p < .001; education, t(1,446) = −4.81, p < .001; subjective health, t(1,401) = −9.38, p < .001; depression symptoms, t(1,426) = 3.39, p = .001; and religious service attendance, t(1478) = −4.63, p < .001, but not in terms of religious commitment, t(1382) = −0.47, p = .636; religious doubt, t(1,318) = −1.31, p = .189; positive religious coping, t(1,358) = −0.47, p = .638; or negative religious coping, t(1,260) = 1.82, p = .069. The mean number of observations per individual was 3.6 (SD = 0.9). To control for the design of the sample, weights were applied both to the GMM analyses and to the logistic regression analyses described subsequently.

The GMM analysis for positive religious coping indicated that the optimal solution was provided by a model including five latent growth classes. Consistent with suggestions for applying GMM (Jung & Wickrama, 2008), we compared the fit with the data for models with successively larger numbers of latent classes. The five-class model provided better fit than a four-class model (Bayesian information criterion [BIC] = 17,571 vs. BIC = 17,622), whereas the six-class model produced some classes with small Ns and had relatively low-average probabilities for latent class membership in some classes and thus the 5-class model was selected. The same procedure was used to identify an optimal structure of three latent classes for growth curves of negative religious coping (three-class BIC = 16,908, 2-class BIC = 16,929, and 4-class GMM produced at least one class with N < 1% of the sample).

The mean intercept and linear slope coefficients within each class are reported in Table 1. Figure 1 illustrates the plot of the mean trajectories in each class for positive religious coping across the course of the older adult age range. Classes 1, 2, and 3 are all characterized by relatively high levels of positive coping early in this range. Although older adults in Class 1 (n = 789) maintain stable levels of positive religious coping as they get older, those in Classes 2 and 3 exhibit mean trajectories of decreasing positive coping, with Class 2 (n = 178) declining relatively slowly and modestly and Class 3 (n = 31) declining much more steeply. Classes 4 and 5 are both characterized by relatively low levels of positive religious coping early on. Older adults in Class 4 (n = 49) showed significant increases in the use of positive religious coping as they got older, whereas those in Class 5 (n = 27) maintained stable low levels of this form of coping.

Table 1.

Summary of Latent Growth Trajectories From Growth Mixture Models for Positive and Negative Religious Coping.

	Intercept	Slope	n (%)
	b (SE)	b (SE)	n (%)
Positive religious coping
Class 1	18.6*** (0.06)	0.003 (0.008)	789 (73.4)
Class 2	15.6*** (0.2)	−0.07** (0.02)	178 (16.6)
Class 3	11.3*** (0.9)	−0.3** (0.1)	31 (2.9)
Class 4	12.3*** (0.4)	0.4*** (0.06)	49 (4.5)
Class 5	6.3*** (0.6)	−0.06 (0.06)	27 (2.5)
Negative religious coping
Class 1	9.7*** (0.4)	0.3*** (0.07)	29 (2.7)
Class 2	6.2*** (0.8)	−0.02* (0.007)	995 (92.5)
Class 3	10.5*** (0.3)	−0.4*** (0.04)	51 (4.8)

Note. SE = standard error.

***p < .001. **p < .01. *p < .05.

Figure 1.

Mean age-related trajectories of positive religious coping by class.

Figure 2 illustrates mean growth curves for the negative religious coping classes. Most of the sample fell into Class 2 (n = 995), which was characterized by low and relatively stable negative religious coping (there was a small but significant average trajectory of decreasing use of this form of coping). A much smaller proportion of older adults were categorized in Class 1 (n = 29) or Class 3 (n = 51), which underwent strong patterns of increasing and decreasing negative religious coping, respectively.

Figure 2.

Mean age-related trajectories of negative religious coping by class.

We examined between-class differences in the number of available observations per individual, as well as baseline age, to evaluate possible sources of bias. The only significant difference that emerged was for observations per individual by positive religious coping trajectory class (p = .040), which indicated that individuals in Class 5 (stable low-positive coping) had slightly more complete observations on average than those in Class 1 (stable high-positive coping), where Class 1 had a mean of 3.6 (SD = 0.9) and Class 5 had a mean of 4.1 (SD = 1.0). Given that both groups had a relatively large number of mean observations and both had stable trajectories, it seems unlikely that this difference would have contributed to any bias in estimation. There were no differences in mean observations among negative coping trajectory classes nor were there differences on either set of GMM classes for baseline age (see Table 2).

Table 2.

Descriptive Statistics Within Latent Positive Religious Coping Classes.

	Class 1 Stable High	Class 2 Moderate Decrease	Class 3 Rapid Decrease	Class 4 Rapid Increase	Class 5 Stable Low	p
Observations per individual	3.6^a (0.9)	3.6^a,b (0.9)	3.9^a,b (0.8)	3.6^a,b (1.0)	4.1^b (1.0)	.040
Age	73.8^a (5.9)	74.9^a (6.3)	72.6^a (6.0)	73.2^a (.2)	74.0^a (6.0)	.128
Race						<.001
White	337 (59.1%)	139 (24.4%)	28 (4.9%)	40 (7.0%)	26 (4.6%)
Black	453 (89.4%)	40 (7.9%)	4 (0.8%)	9 (1.8%)	1 (0.2%)
Gender						<.001
Male	279 (64.1%)	77 (17.7%)	19 (4.4%)	39 (9.0%)	21 (4.8%)
Female	517 (79.8%)	103 (15.9%)	12 (1.9%)	10 (1.5%)	6 (0.9%)
Education (years)	11.5^a (3.3)	11.9^a,b (3.4)	13.3^b (3.6)	12.1^a,b (3.1)	14.4^c (3.0)	<.001
Subjective health	7.3^a (1.5)	7.4^a (1.5)	7.8^a,b (1.5)	7.2^a (1.5)	8.1^b (1.1)	.012
Depression symptoms	4.2^a (4.8)	3.9^a (4.3)	4.4^a (5.9)	4.2^a (4.1)	3.5^a (3.2)	.918
Religious factors
Attendance	6.5^a (2.4)	5.4^b (2.6)	4.1^c (2.3)	3.1^c,d (2.2)	1.8^d (1.7)	<.001
Commitment	10.6^a (1.5)	9.2^b (1.5)	8.7^c (2.3)	7.3^c,d (1.8)	6.1^d (2.5)	<.001
Doubt	6.3^a (2.3)	7.5^b (2.7)	9.5^c (3.6)	10.9^c,d (3.5)	12.9^d (5.5)	<.001
Denomination						<.001
Catholic	116 (59.5%)	62 (31.8%)	6 (3.1%)	11 (5.6%)	0 (0.0%)
Mainline	110 (58.8%)	40 (21.7%)	15 (8.2%)	14 (7.6%)	5 (2.7%)
Evangelical	446 (86.4%)	43 (8.3%)	7 (1.4%)	13 (2.5%)	5 (1.4%)
Other	57 (76.0%)	10 (13.3%)	2 (2.7%)	4 (5.3%)	2 (2.7%)
Religious coping (baseline)
Positive	18.7^a (1.5)	15.5^b (2.7)	15.0^b (3.4)	9.2^c (2.3)	6.3^d (1.9)	<.001
Negative	8.6^a (3.1)	8.3^a (3.3)	9.0^a (4.1)	8.2^a (2.9)	5.8^b (1.0)	.002
Negative coping trajectory						.475
Class 1	20 (69.0%)	6 (20.7%)	0 (0.0%)	2 (6.9%)	1 (3.4%)
Class 2	733 (73.1%)	171 (17.0%)	29 (2.9%)	44 (4.4%)	26 (2.6%)
Class 3	42 (80.8%)	3 (5.8%)	3 (5.8%)	3 (5.8%)	1 (1.9%)

Note. All analyses are weighted. Significance levels reflect χ² tests for continuous variables (race, gender, denomination, and negative coping class) and analysis of variance for all others. Superscripts denote homogeneous subsets with overlapping 95% confidence intervals.

In the next phase of the analysis, we examined key demographic and religious factors related to membership in these latent growth classes. The GMM procedure produces predicted class membership for each individual in the sample. These estimates were merged into the Wave 1 survey data, allowing us to examine the association between baseline variables and subsequent coping trajectories. Table 2 provides descriptive statistics for each positive coping class. A number of between-class differences emerged. In particular, high and stable use of positive religious coping (Class 1) was much more prevalent among African Americans than among White participants, with 89.4% of the African American sample being classified in this group, compared to 59.1% of the White sample. There was also a substantial difference by gender, with 79.8% of women but only 64.1% of men being grouped in Class 1. In terms of other demographic and health-related factors, the average level of education was highest among those whose use of positive religious coping was low and stable throughout the older adult life course (Class 5) and lowest among those with high and stable coping (Class 1). Similarly, subjective health was highest in Class 2 and lowest in Class 1. In terms of religious characteristics, Evangelicals were considerably more likely to be in the high and stable group than members of other denominations (86.4% of evangelicals, in comparison to 59.5% of catholics and 58.8% of mainline protestants). Members of the high and stable trajectory class tended to report higher religious attendance and religious commitment, and lower levels of religious doubt, in comparison to other categories. Baseline levels of positive religious coping were strongly associated with trajectory class, and coping levels tended to remain stable in those with the lowest and highest baseline scores (Class 5 and Class 1, respectively), whereas the classes exhibiting change over time (Classes 2, 3, and 4) had more moderate and more widely variable baseline levels of positive coping. By contrast, there was little difference in baseline negative religious coping between classes, with the exception of Class 5 (low and stable positive coping), which also exhibited much lower levels of negative religious coping at baseline. There was no association between positive and negative latent class membership.

Corresponding descriptive statistics by negative religious coping category are provided in Table 3. There were relatively few group differences, in comparison to those observed among the positive religious coping categories. Both depression symptoms and doubt were significantly lower at baseline in the stable negative coping category (Class 2) as was the initial extent of negative religious coping itself. There were no group differences in race, gender, education, subjective health, and attendance at religious services, religious commitment, religious denomination, or positive coping trajectory class.

Table 3.

Descriptive Statistics Within Latent Negative Religious Coping Classes.

	Class 1 Rapid Increase	Class 2 Slight Decrease	Class 3 Rapid Decrease	p
Observations per individual	4.0^a (0.8)	3.6^a (0.9)	3.7^a (0.9)	.166
Age	74.2^a (6.5)	74.0^a (6.0)	72.7^a (6.0)	.295
Race				.333
White	13 (2.3%)	533 (93.5%)	24 (4.2%)
Black	17 (3.3%)	463 (91.1%)	28 (5.5%)
Gender				.335
Male	10 (2.3%)	409 (91.7%)	16 (3.7%)
Female	19 (2.9%)	594 (91.7%)	35 (5.4%)
Education (years)	11.1^a (3.3)	11.8^a (3.4)	10.9^a (3.1)	.174
Subjective health	7.2^a (1.3)	7.3^a (1.5)	6.9^a (1.5)	.091
Depression symptoms	5.5^a,b (5.4)	4.0^a (4.5)	6.3^b (6.1)	<.001
Religious factors
Attendance	6.6^a (2.2)	6.0^a (2.6)	6.1^a (2.7)	.424
Commitment	10.0^a (2.0)	10.1^a (1.8)	10.1^a (1.8)	.992
Doubt	8.4^a (3.5)	6.8^b (2.8)	8.1^a (3.8)	<.001
Denomination				.602
Catholic	4 (2.1%)	182 (93.8%)	8 (4.1%)
Mainline	7 (3.8%)	168 (90.8%)	10 (5.4%)
Evangelical	15 (2.9%)	472 (91.7%)	28 (5.4%)
Other	1 (1.3%)	73 (97.3%)	1 (1.3%)
Religious coping (baseline)
Positive	17.1^a (3.9)	17.4^a (3.4)	17.7^a (3.8)	.780
Negative	10.8^a (4.4)	8.3^b (2.9)	12.8^a (4.7)	<.001
Positive coping trajectory				.475
Class 1	20 (2.5%)	733 (92.2%)	42 (5.3%)
Class 2	6 (3.3%)	171 (95.0%)	3 (1.7%)
Class 3	0 (0.0%)	29 (90.6%)	3 (9.4%)
Class 4	2 (4.1%)	44 (89.8%)	3 (6.1%)
Class 5	1 (3.6%)	26 (92.9%)	1 (3.6%)

Note. All analyses are weighted. Significance levels reflect χ² tests for continuous variables (race, gender, denomination, and positive coping class) and analysis of variance for all others. Superscripts denote homogeneous subsets with overlapping 95% confidence intervals.

Finally, we used nominal logistic regression analysis to examine baseline factors predicting the subsequent trajectory of positive and negative religious coping. These models include the variables described earlier, with two exceptions: Because of the low variability in class distribution among African Americans and Evangelicals (a very high proportion of whom were in the high and stable class, with very small numbers in Classes 3 to 5 in particular), the models did not reliably converge when race or religious denomination were entered and thus they were excluded. Baseline religious coping was included to control for the interrelatedness of religious coping and other religious factors.

Results for positive religious coping categories are given in Table 4. Compared with being part of the high and stable positive religious coping trajectory category (Class 1), older adults were more likely to go on to exhibit a pattern of moderately decreasing coping (Class 2) if they had low baseline religious commitment (b = −0.26, p < .001, odd ratios [OR] = 0.78). They were more likely to subsequently experience rapid decrease (Class 3) if they were in better health (b = 0.43, p = .010, OR = 1.54), attended religious services less frequently (b = −0.25, p = .004, OR = 0.78), and had higher religious doubt (b = 0.23, p = .002, OR = 1.25). Rapid increases (Class 4) were more likely for men than for women (b = 1.29, p = .014, OR = 3.63), for those with less frequent religious attendance (b = −0.23, p = .027, OR = 0.79) and higher religious doubt (b = 0.23, p = .009, OR = 1.26). They were more likely to have stable low-positive religious coping (Class 5), if they had higher levels of education (b = 0.28, p = .029, OR = 1.32), lower religious attendance (b = −0.54, p = .005, OR = 0.58), and higher religious doubt (b = 0.23, p = .029, OR = 1.25).

Table 4.

Nominal Logistic Regression Results for Positive Religious Coping Trajectory Class (Odds Ratios and 95% Confidence Intervals).

	Class 2 Moderate Decrease	Class 3 Rapid Decrease		Class 4 Rapid Increase	Class 5 Stable Low
	Vs. Class 1	Vs. Class 1	Vs. Class 2	Vs. Class 1	Vs. Class 1	Vs. Class 4
Baseline coping	0.60*** [0.54, 0.66]	0.60*** [0.50, 0.71]	1.01 [0.86, 1.17]	0.39*** [0.31, 0.49]	0.29*** [0.20, 0.42]	0.75* [0.58, 0.97]
Gender (Male)	0.89 [0.59, 1.34]	1.75 [0.77, 3.98]	1.96 [0.87, 4.43]	3.63* [1.29, 10.19]	2.14 [0.46, 10.01]	0.59 [0.14, 2.44]
Education	1.01 [0.94, 1.07]	1.14 [1.00, 1.30]	1.13 [0.99, 1.29]	1.02 [0.88, 1.19]	1.32* [1.03, 1.70]	1.29* [1.03, 1.62]
Health	1.12 [0.96, 1.30]	1.54* [1.11, 2.14]	1.38 [1.00, 1.92]	1.04 [0.73, 1.49]	1.39 [0.78, 2.51]	1.34 [0.81, 2.21]
Depression symptoms	0.98 [0.94, 1.03]	1.02 [0.94, 1.13]	1.05 [0.95, 1.15]	0.98 [0.87, 1.09]	0.98 [0.81, 1.19]	1.01 [0.85, 1.19]
Religious factors
Attendance	0.98 [0.88, 1.05]	0.78** [0.66, 0.92]	0.81* [0.69, 0.96]	0.79* [0.65, 0.98]	0.58** [0.40, 0.85]	0.73 [0.52, 1.04]
Commitment	0.78*** [0.67, 0.90]	0.78 [0.58, 1.03]	1.00 [0.76, 1.33]	0.70 [0.49, 1.01]	0.66 [0.42, 1.03]	0.94 [0.68, 1.30]
Doubt	1.07 [0.98, 1.17]	1.25** [1.09, 1.44]	1.17* [1.02, 1.35]	1.26** [1.06, 1.51]	1.25* [1.02, 1.54]	0.99 [0.85, 1.15]

***p < .001. **p < .01. *p < .05.

In addition to the main model, in which Class 1 was the comparison category, we examined baseline factors differentiating (a) individuals with initially relatively high levels of positive coping who went to experience rapid versus moderate decreases (Class 2 vs. Class 3) and (b) individuals with initially lower levels of positive coping who went on to either increase or remain stable (Class 4 vs. Class 5). Rapid decrease verues moderate decrease was associated with lower attendance (b = −0.21, p = .015, OR = 0.81) and higher religious doubt (b = 0.16, p = .025, OR = 1.17). Stable low-positive coping versus low but increasing coping was associated only with higher education (b = 0.26, p = .026, OR = 1.29).

Logistic regression results for negative religious coping trajectory categories are given in Table 5. In comparison to low and stable negative coping (Class 2), a pattern of increasing negative coping (Class 1) was related to more frequent religious attendance (b = 0.20, p = .041, OR = 1.23) and greater religious doubt (b = 0.20, p = .001, OR = 1.22). A pattern of decreasing negative coping (Class 3) was related to greater religious doubt only (b = 0.17, p = .001, OR = 1.19). No baseline model variables significantly differentiated between increasing and decreasing trajectories.

Table 5.

Nominal Logistic Regression Results for Negative Religious Coping Trajectory Class (Odds Ratios and 95% Confidence Intervals).

	Class 1	Class 3
	Vs. 2	Vs. 2	Vs. 1
Baseline coping	1.02 [0.89, 1.20]	1.05 [0.93, 1.19]	1.04 [0.85, 1.26]
Gender (Male)	0.85 [0.38, 1.92]	0.75 [0.40, 1.41]	0.88 [0.32, 2.38]
Education	0.91 [0.81, 1.03]	0.92 [0.83, 1.01]	1.00 [0.87, 1.16]
Health	1.04 [0.79, 1.36]	0.92 [0.75, 1.13]	0.89 [0.64, 1.23]
Depression symptoms	1.06 [0.99, 1.15]	1.07* [1.01, 1.13]	1.00 [0.92, 1.10]
Religious factors
Attendance	1.23* [1.01, 1.49]	1.10 [0.96, 1.26]	0.90 [0.71, 1.13]
Commitment	0.97 [0.74, 1.27]	0.98 [0.80, 1.20]	1.01 [0.73, 1.40]
Doubt	1.22** [1.08, 1.37]	1.19** [1.08, 1.31]	0.98 [0.85, 1.12]

***p < .001. **p < .01. *p < .05.

Discussion

The results of this study indicate that, consistent with the aged heterogeneity perspective (Nelson & Dannefer, 1992), there is substantial heterogeneity in the course of individual changes in the use of religious coping during older adulthood, at least in this largely Christian U.S. sample. The nature of these changes was only partially in keeping with our hypotheses. Although positive coping was expected to increase for most older adults (Hypothesis 1a), as a consequence of changing needs and increasing salience of religious concerns (Erikson, 1968; Tornstam, 1997; Vail et al., 2010), only a small minority of older adults followed this pattern, with the majority exhibiting high but stable levels of positive religious coping during the course of the study. However, the expectation that negative religious coping would tend to decrease during older adulthood (Hypothesis 1b) appeared to hold true for the vast majority of the sample. Consistent with the aged heterogeneity perspective (Nelson & Dannefer, 1992), distinct classes of age-related trajectories of change were detected for both positive (Hypothesis 2a) and negative (Hypothesis 2b) religious coping. Finally, our analyses of which baseline factors helped predict which trajectory each older adult would subsequently follow were consistent with the expectations that demographics (Hypothesis 3), stressors (Hypothesis 4), and religious background (Hypothesis 5) would be associated with patterns of change.

These findings suggest that the story of the course of religious coping in late life is more nuanced than cross-sectional analyses comparing older adults with other age-groups might indicate. Although these results recapitulate the broad finding that older adults tend to practice relatively high levels of positive religious coping, and relatively low levels of negative religious coping, on average, there is important variation between individuals not only in terms of the overall extent of religious coping but also in terms of changes in coping. This is particularly true in the case of positive religious coping. More than a quarter of older adults did not follow the “typical” trajectory of high and stable positive religious coping, with the majority of these experiencing substantial declines over the course of late life. Fewer older adults deviated from the typical pattern with respect to changes in negative religious coping, but the number was nevertheless significant, and there was a pattern of significant mean change in every trajectory class. Taken together, these findings lend support to the picture of older adulthood as a time of both change and diversification, with coping on these dimensions becoming less homogenous in the decades after the age of 65 years.

Equally important is the analysis of the antecedents of these changes. There were a number of informative results with respect to positive religious coping, whereas those regarding negative coping remain somewhat elusive. Substantial differences were observed by race and gender, with African Americans and women more likely to have high and stable patterns of positive religious coping in comparison to Whites and men. Religious involvement and education were the most important predictors of specific patterns of change in coping. It is possible to divide these patterns into two general groups based on levels of religious coping at the beginning of the older adulthood: those with initially high levels (who subsequently either remained stable, decreased moderately, or decreased rapidly) and those with initially low levels (who subsequently either remained stable or increased rapidly). Compared with high and stable coping, moderate declines were associated with lower baseline religious commitment, whereas rapid declines were associated with a number of factors including low religious attendance, high religious doubt, good health, and high education. Among those with initially low religious coping use, only education significantly predicted stability versus change, with lower education being related to a pattern of increasing religious coping.

These findings appear to argue against a universal developmental pattern of increasing religious salience in late life. Specifically, marginally affiliated older adults seem to move away from the use of religion as a coping resource, with smaller declines among those with lower personal commitment (but who still have relatively high involvement with the social aspects of religion associated with service attendance and who are relatively untroubled by religious doubt), and the largest declines among those who are disengaged with the religious group and who experience higher levels of doubt about their religious beliefs. The role of education may indicate that older adults with a stronger secular worldview may have a greater array of coping options, particularly when it comes to cognitive strategies such as reframing and meaning making (Vail et al., 2010), lessening their reliance on religious coping. Likewise, older adults with initially low religious engagement may begin to seek out religious coping strategies when they have less exposure to secular alternatives for making meaning, offering one possible explanation for the relationship between lower education and rapid increases in positive religious coping within this subgroup.

When it comes to changes in negative religious coping, the only factor found in this study to predict differences in trajectory was religious doubt. Interestingly, baseline doubt was elevated not only among those older adults who subsequently engaged in more negative coping but also in those whose negative coping declined rapidly over time. Religious uncertainty may tend to contribute to a generally volatile pattern of thinking about religious matters, which may influence broad swings in negative coping. Alternatively, salience of some of the disturbing or contradictory elements of religion brought up by negative coping use may contribute to religious doubts. Furthermore, there individuals may react differently to the experience of doubt. Successful resolution of doubt may lead to personal growth, whereas unsuccessful resolution may lead to despair. This could mean that for those who are able to successfully resolve doubt through personal growth, negative coping responses decline, but for those who are unable to resolve their doubts, more negative coping ensues. More detailed measures of doubt are needed to examine these differences.

These results may help provide insights that can be applied to addressing health risks in older adulthood. Previous research has found that religious coping can be important for late life health outcomes (Pargament, Feuille, & Burdzy, 2011) and thus predicting which individuals will face patterns of reduced positive coping and increased negative coping may help target health interventions aimed at mitigating these risks. As White men are at elevated risk of suicide in older adulthood (Conwell, Duberstein, & Caine, 2002), the finding that these individuals are also the most likely to experience instability in religious coping may be instructive in understanding some of the causes of this problem and may be useful in identifying individuals at highest risk.

Limitations of this study include the relatively limited longitudinal span covered for each individual trajectory. The maximum time in the study was 11 years, with many older adults being represented by a shorter span, due to mortality or attrition from other causes. Although it is possible to use statistical modeling procedures to aggregate these individual trajectories into longer generalized age-related trajectories, this necessarily leaves the possibility that these trajectories may have had some differences due to age-related factors (e.g., cohort effects) that cannot be accounted for in this model due to conflation of age and birth year. That is, the oldest range of the age-related model is based entirely on those born in the earliest cohorts, whereas the youngest end of the age range is constructed entirely from the latest cohorts. Furthermore, it is difficult to assess the potential for quadratic or other curvilinear trajectories when many individuals have only a relatively small number of repeated observations. A longer longitudinal period with sample replacement and a larger initial sample (to overcome the effects of attrition by the oldest ages) are needed to address these issues.

Additionally, the use of the GMM procedure allowed us to examine distinct trajectory classes in a way that a conventional conditional growth curve model would not but also precluded the use of time-varying covariates. Hence, this analysis examines only baseline health and religion factors related to changes in religious coping rather than accounting for mutual change in both across time. Although the relatively high degree of autocorrelation for these predictors within individuals across time helps to justify this approach, the use of time-varying methods would help to support these results. Relatedly, individuals with no longitudinal data (i.e., those lost to attrition after the first wave) were omitted from the analyses. Individuals excluded in this way tended to be older, less educated, in worse physical and mental health, and less frequently attending religious services than the rest of the sample at baseline, which may limit generalizability of these findings among those with the most background risk factors for a variety of problems and thus most in need of coping resources. Finally, the sample was limited to Black and White U.S. individuals with Christian religious backgrounds, and generalizability to those from other cultural contexts needs further research to be established.

As these results help to illustrate, the picture of late life—and particularly that related to the use of religious coping to deal with the stressors of late life—given by cross-sectional studies comparing older and younger adults is incomplete. Rather than being a period of stasis, it is often a time when individuals may undergo considerable change and when the population becomes more heterogeneous and dispersed (Nelson & Dannefer, 1992). From a theoretical standpoint, these results highlight the importance of considering the dynamics of change within the span of older adulthood when studying the relationship between religion and individual development. From a practical standpoint, they suggest that individual differences in religious background and involvement, personal history (e.g., education), and sources of stress may make a difference in the health and well-being benefits that older adults are able to obtain from religious forms of coping (Ano & Vasconcelles, 2005; Harrison et al., 2001). We hope that this study will serve as an initial step to stimulate more research in the domain of late-life heterogeneity in religious coping and other elements of individual religion.

Footnotes

Declaration of Conflicting Interests

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

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by grants from the National Institute on Aging (R01 AG026259) and the John Templeton Foundation (20887).

References

Alferi

S. M.

Culver

J. L.

Carver

C. S.

Arena

P. L.

Antoni

M. H.

(1999). Religiosity, religious coping, and distress: A prospective study of catholic and evangelical Hispanic women in treatment for early-stage breast cancer. Journal of Health Psychology, 4, 343–356. doi:http://doi.org/10.1177/135910539900400304

Ano

G. G.

Vasconcelles

E. B.

(2005). Religious coping and psychological adjustment to stress: A meta-analysis. Journal of Clinical Psychology, 61, 461–480. doi:http://doi.org/10.1002/jclp.20049

Chatters

L. M.

Taylor

R. J.

Jackson

J. S.

Lincoln

K. D.

(2008). Religious coping among African Americans, Caribbean Blacks and non-Hispanic Whites. Journal of Community Psychology, 36, 371–386. doi:http://doi.org/10.1002/jcop.20202

Conwell

Duberstein

P. R.

Caine

E. D.

(2002). Risk factors for suicide in later life. Biological Psychiatry, 52, 193–204. doi:http://doi.org/10.1016/S0006-3223(02)01347-1

Courtenay

B. C.

Poon

L. W.

Martin

Clayton

G. M.

Johnson

M. A.

(1992). Religiosity and adaptation in the oldest-old. The International Journal of Aging and Human Development, 34, 47–56. doi:http://doi.org/10.2190/N058-Y7X6-YLGJ-XBGE

Erikson

E. H.

(1968). Identity, youth, and crisis. New York, NY: W. W. Norton.

Fetzer Institute & National Institute on Aging. (1999). Multidimensional measurement of religiousness/spirituality for use in health research. Kalamazoo, MI: John E. Fetzer Institute.

Graham

Olchowski

Gilreath

(2007). How many imputations are really needed? Some practical clarifications of multiple imputation theory. Prevention Science, 8, 206–213. doi:http://doi.org/10.1007/s11121-007-0070-9

Harrison

M. O.

Koenig

H. G.

Hays

J. C.

Eme-Akwari

A. G.

Pargament

K. I.

(2001). The epidemiology of religious coping: A review of recent literature. International Review of Psychiatry, 13, 86–93. doi:http://doi.org/10.1080/09540260120037317

10.

Hay

E. L.

Diehl

(2010). Reactivity to daily stressors in adulthood: The importance of stressor type in characterizing risk factors. Psychology and Aging, 25, 118–131.

11.

Hayward

R. D.

Krause

(2013). Patterns of change in religious service attendance across the life course: Evidence from a 34-year longitudinal study. Social Science Research, 42, 1480–1489. doi:http://doi.org/10.1016/j.ssresearch.2013.06.010

12.

Hoare

C. H.

(2002). Erikson on development in adulthood: New insights from the unpublished papers. New York, NY: Oxford University Press.

13.

Horton

N. J.

Lipsitz

S. R.

(2001). Multiple imputation in practice. The American Statistician, 55, 244–254. doi:http://doi.org/10.1198/000313001317098266

14.

Jung

Wickrama

K. A. S.

(2008). An introduction to latent class growth analysis and growth mixture modeling. Social and Personality Psychology Compass, 2, 302–317. doi:http://doi.org/10.1111/j.1751-9004.2007.00054.x

15.

Koenig

H. G.

Pargament

K. I.

Nielsen

(1998). Religious coping and health status in medically ill hospitalized older adults. Journal of Nervous and Mental Disease, 186, 513–521.

16.

Krause

(2002). A comprehensive strategy for developing closed-ended survey items for use in studies of older adults. The Journals of Gerontology Series B: Psychological Sciences and Social Sciences, 57, S263–S274. doi:http://doi.org/10.1093/geronb/57.5.S263

17.

Krause

(2008). Aging in the church: How social relationships affect health. West Conshohocken, PA: Templeton Foundation Press.

18.

Levin

J. S.

Taylor

R. J.

Chatters

L. M.

(1994). Race and gender differences in religiosity among older adults: Findings from four national surveys. Journal of Gerontology: Social Sciences, 49, S137–S145. doi:http://doi.org/10.1093/geronj/49.3.S137

19.

McEwen

B. S.

(2008). Central effects of stress hormones in health and disease: Understanding the protective and damaging effects of stress and stress mediators. European Journal of Pharmacology, 583, 174–185. doi:http://doi.org/S0014-2999(08)00027-7 [pii] 10.1016/j.ejphar.2007.11.071

20.

Mroczek

D. K.

Almeida

D. M.

(2004). The effect of daily stress, personality, and age on daily negative affect. Journal of Personality, 72, 355–378. doi:http://doi.org/10.1111/j.0022-3506.2004.00265.x

21.

Muthén

L. K.

(2000). Integrating person-centered and variable-centered analyses: Growth mixture modeling with latent trajectory classes. Alcoholism: Clinical and Experimental Research, 24, 882–891. doi:http://doi.org/10.1111/j.1530-0277.2000.tb02070.x

22.

Nelson

E. A.

Dannefer

(1992). Aged heterogeneity: Fact or fiction? The fate of diversity in gerontological research. The Gerontologist, 32, 17–23. doi:http://doi.org/10.1093/geront/32.1.17

23.

Pargament

K. I.

(1997). The psychology of religion and coping: Theory, research, and practice. New York, NY: Guilford.

24.

Pargament

K. I.

Feuille

Burdzy

(2011). The brief RCOPE: Current psychometric status of a short measure of religious coping. Religions, 2, 51–76.

25.

Pargament

K. I.

Koenig

H. G.

Perez

L. M.

(2000). The many methods of religious coping: Development and initial validation of the RCOPE. Journal of Clinical Psychology, 56, 519–543. doi:http://doi.org/10.1002/(sici)1097-4679(200004)56:4<519::aid-jclp6>3.0.co;2-1

26.

Pargament

K. I.

Koenig

H. G.

Tarakeshwar

Hahn

(2004). Religious coping methods as predictors of psychological, physical and spiritual outcomes among medically ill elderly patients: A two-year longitudinal study. Journal of Health Psychology, 9, 713–730. doi:http://doi.org/10.1177/1359105304045366

27.

Park

C. L.

Cohen

L. H.

(1993). Religious and nonreligious coping with the death of a friend. Cognitive Therapy and Research, 17, 561–577. doi:http://doi.org/10.1007/BF01176079

28.

Pew Forum on Religion and Public Life. (2012). “Nones” on the rise: One-in-five adults have no religious affiliation. Washington, DC: Pew Research Center.

29.

Radloff

L. S.

(1977). The CES-D scale. Applied Psychological Measurement, 1, 385–401. doi:http://doi.org/10.1177/014662167700100306

30.

Schuster

M. A.

Stein

B. D.

Jaycox

L. H.

Collins

R. L.

Marshall

G. N.

Elliott

M. N.

… Berry

S. H.

(2001). A national survey of stress reactions after the September 11, 2001, terrorist attacks. New England Journal of Medicine, 345, 1507–1512. doi:http://doi.org/10.1056/NEJM200111153452024

31.

Schwadel

(2011). Age, period, and cohort effects on religious activities and beliefs. Social Science Research, 40, 181–192. doi:http://doi.org/10.1016/j.ssresearch.2010.09.006

32.

Steensland

Park

J. Z.

Regnerus

M. D.

Robinson

L. D.

Wilcox

W. B.

Woodberry

R. D.

(2000). The measure of American religion: Toward improving the state of the art. Social Forces, 79, 291–318. doi:http://doi.org/10.2307/2675572

33.

Tepper

Rogers

S. A.

Coleman

E. M.

Malony

H. N.

(2001). The prevalence of religious coping among persons with persistent mental illness. Psychiatric Services, 52, 660–665. doi:http://doi.org/10.1176/appi.ps.52.5.660

34.

Tornstam

(1997). Gerotranscendence: The contemplative dimension of aging. Journal of Aging Studies, 11, 143–154. doi:http://doi.org/10.1016/S0890-4065(97)90018-9

35.

U.S. Census Bureau. (2012). 2012 National Population Projections. Retrieved from http://www.census.gov/population/projections/data/national/2012.html

36.

Vail

K. E.

Rothschild

Z. K.

Weise

D. R.

Solomon

Pyszczynski

Greenberg

(2010). A terror management analysis of the psychological functions of religion. Personality and Social Psychology Review, 14, 84–94. doi:http://doi.org/10.1177/1088868309351165