Full text of DOI:10.1093/geronb/56.6.S326

Journal of Gerontology: SOCIAL SCIENCES
Copyright 2001 by The Gerontological Society of America
2001, Vol. 56B, No. 6, S326–S334
Patterns of Religious Practice and Belief
in the Last Year of Life
Ellen L. Idler,1 Stanislav V. Kasl,2 and Judith C. Hays3
1Institute for Health, Health Care Policy, and Aging Research, Rutgers University, New Brunswick, New Jersey.
2Department of Epidemiology and Public Health, Yale University School of Medicine, New Haven, Connecticut.
3Department of Psychiatry, Duke University Medical Center, Durham, North Carolina.
Objectives. Although it is frequently assumed that issues of religious faith become more salient at the end of life,
there is little or no population-based empirical evidence testing this assumption.
Methods. Using data from the New Haven site of the Established Populations for Epidemiologic Studies of the Elderly
(N ꢀ 2,812), the authors examined self-reports of attendance at services, self-ratings of religiousness, and strength and
comfort felt from religion for respondents who did and did not die within 12 months following an interview. Religious-
ness was assessed at baseline (1982) and in follow-up interviews in 1985, 1988, and 1994. Cross-sectional comparisons
of levels of religiousness were made among persons in their last 6 months of life, persons in their last 12 months of life,
and persons who survived 12 months, and longitudinal comparisons were made with religiousness at the previous wave.
Results. After adjusting for age, sex, education, marital status, religious affiliation, and a set of health status mea-
sures, the authors found that although attendance at religious services declined among the near-deceased, this group
showed either stability or a small increase in feelings of religiousness and strength/comfort received from religion.
Overall levels of attendance and religious feelings were high for this religiously diverse sample.
Discussion. Community studies of respondents in their last year of life are rare. In this sample, religious involve-
ment appears to continue throughout the last months of life.
CCORDING to conventional wisdom about philoso-
phers’ wagers and the absence of atheists in foxholes,
and practices with their own responses to these questions 3
years previously. The large size, religious diversity, and
representativeness of the sample; the prospective design of
the study; and the availability of both behavioral and sub-
jective measures of religious involvement make this study
unique in the literature on the end of life.
A
religious faith is of critical importance at the end of life.
But how salient are issues of faith and spirituality in the pe-
riod immediately preceding death? Are there identifiable
changes at the end of life, signifying a common experience
of these issues in old age? Or do religious beliefs and prac-
tices remain stable throughout this stage of life? We know
very little generally about the quality of life at the end of life
and practically nothing about religious and spiritual identi-
ties in this critical period (Bradley, Fried, Kasl, & Idler,
2000). A recent review of the literature on ethics and spiri-
tuality at the end of life concluded that although a great deal
has been written about ethics, very little has been written
about spirituality (Bevins & Cole, 2000). Moreover, the
small number of studies of religiousness at the end of life
are primarily in the nursing literature and are often limited
to populations of AIDS or cancer patients, many of them
relatively young.
In this study we had a rare opportunity to examine the
patterns of religious belief and practice in a large and repre-
sentative sample of elderly people in their last year of life.
The sample was composed of healthy elderly people living
independently in a religiously pluralistic community, and
observed over 14 years. With repeated observations of reli-
gious practice and belief available to us, we compared those
in their last 6 and 12 months of life to respondents who
survived 12 months or longer following an interview; we
also compared last-year-of-life respondents’ religious beliefs
Expectations of Decline
Theory in social gerontology is equivocal with respect to
religion and aging. On the one hand, the early and now dis-
regarded disengagement theory suggests that individuals
should, and do, withdraw from social roles, including reli-
gious roles, as they reach old age (Cumming & Henry,
1961). Cumming and Henry argued that both religious piety
and attendance at religious services should and did decline
sharply among their respondents aged older than 75. Two
decades later, the disengagement theory again became the
basis for expectations about religious involvement in old
age; Mindel and Vaughan (1978), in a small but influential
study, argued that disengagement from religion appeared to
take place with some forms of religious involvement but not
others. In their sample of 106 elderly Missourians, they ob-
served that levels of organizational religious activity were
not strongly correlated with nonorganizational religious ac-
tivity; nonattenders were just as likely as attenders to listen
to religious services or music on radio or television or to be-
lieve that their religion helped them understand their lives.
The health of the respondents was a strong influence on
these differences; when compared with the unimpaired
S326

RELIGIOUSNESS IN THE LAST YEAR OF LIFE
S327
group, the impaired group had lower levels of organiza-
tional religious activities, but higher levels of nonorganiza-
tional activities. The shortcomings of the study, however,
were its cross-sectional design and lack of multivariable
methods. Subsequent studies have shown that although
there is considerable empirical evidence that attendance at
religious services does decline with age (Blazer & Palmore,
1976; Ploch & Hastings, 1994), such declines are primarily
attributable to illness and disability (Idler & Kasl, 1997b;
Levin & Markides, 1985).
Recently, Hays and colleagues (1998) examined one of
the forms of nonorganizational religious activity, the “elec-
tronic church.” They found in a longitudinal study of elderly
North Carolinians, that declines in attendance were caused
primarily by declines in functional ability. They also found
that the use of religious media increased over time, although
this increase was not limited to those with deteriorating
health. Thus the modified view is that there may be disen-
gagement from some forms of religious organization but not
others, and in fact quite high levels of nonorganizational re-
ligious activity may be maintained among impaired respon-
dents.
A second line of research that might lead one to expect
declining religious involvement, and that bears more di-
rectly on the very end of life, is the work regarding terminal
drop or terminal decline (Berg, 1996; Kleemeier, 1962; Sie-
gler, 1975). The terminal decline hypothesis states that
changes in many functions, including but not limited to
higher intellectual functioning, are not correlated primarily
with chronological age (distance from birth), but rather are a
function of the individual’s distance from death. Recent
work in this area has taken a broad view that decline is per-
vasive in many areas of life in the period before death, and
this broad decline represents a diminishing of the vitality of
the self in all areas of functioning, not just in the intellectual
arena. Findings from the Berlin Aging Study (BASE) have
shown that time to death was associated with lower life sat-
isfaction, lower positive affect, and more social loneliness
(Maier & Smith, 1996). No work of which we are aware has
examined religiousness or spirituality in connection with
terminal decline, even though the abstract nature of reli-
gious symbols makes it an intriguing question. To the de-
gree that spiritual capacities or religious beliefs are associ-
ated with higher intellectual functions and/or emotional
equilibrium, in addition to physical functioning, then one
might expect decline in them.
opment culminate in old age, when individuals are faced
with the challenge of looking back over their lives and con-
structing an interpretation of them. Erikson called the psy-
chosocial crisis of old age the crisis of ego integrity versus
despair; the achievement of a sense of individual integrity
he defined as “the ego’s accrued assurance of its proclivity
for order and meaning . . . the acceptance of one’s one and
only life cycle and of the people who have become signifi-
cant to it as something that had to be” (1968, p. 139). In his
later writings, Erikson (Erikson, Erikson, & Kivnick, 1986)
speculated that it might be necessary to add a ninth stage to
his scheme, consisting of “a sense or premonition of immor-
tality . . . as creatively given form in the world religions”
(1986, p. 336–337). Similarly, there is Jung’s blunt and fre-
quently quoted statement that “among all my patients in the
second half of life . . . there has not been one whose problem
in the last resort was not that of finding a religious outlook
on life” (Jung, 1933, p. 229). By and large, however, theo-
ries of psychological development in late life have had
rather little to say about the role of religious beliefs. Inter-
estingly, theories of faith development (Fowler, 1981),
which address issues of the development of religion and
spirituality over the life course, posit at most a rather loose
relationship between age and faith. These theories propose
that development can become arrested at any stage or age,
and that “mature” faith is achieved by a relative few. In our
study we did not have any indicators of maturity of faith.
However, one may expect that, as the individual’s personal
story takes shape and approaches completion, it can be in-
creasingly articulated within the “master story” of the indi-
vidual’s faith tradition, allowing the dying person to find
meaning in his or her life up until that point.
More recent studies of aging and faith have emphasized
the importance of religious coping resources for meeting the
particular challenges of old age such as bereavement, social
isolation, cognitive impairment, pain, disability, and deci-
sions about end-of-life care (Bevins & Cole, 2000; Kimble,
McFadden, Ellor, & Seeber, 1995). Religious coping re-
sources are seen as lowering depression and reducing anxi-
ety, alcohol abuse, and even suicide, all of which are rela-
tively common in elderly populations (Koenig, 1995).
Particularly notable is the work of Pargament (1997) and
colleagues, who have argued that the most challenging mo-
ments of life occur in old age: “Painful as they are, these
critical moments offer a remarkably open window into hu-
man character, faith, and the process of coping with the fi-
nal transitions of life” (Pargament, Van Haitsma, & Ensing,
1995, p. 47). In this view, religious coping resources offer
the elderly individual the ability to maintain and integrate
the self, to retain feelings of mastery over the environment,
to continue their intimate relations with others, and to come
to grips with mortality (Pargament et al., 1995). The success
or failure of religious coping mechanisms is somewhat be-
yond the scope of this article, however. From the coping
perspective we simply drew the premise that religious cop-
ing responses should be especially prominent in old age be-
cause this is the time when existential or health crises are
likely to occur. Operationalizations of “religious coping” in
the research literature have long included subjective states,
cognitions, or emotions, as well as religious practices such
It is also possible that religious faith takes on a different
and negative meaning at the end of life. The passage of time
may eliminate the opportunity for an individual to make res-
titution or receive forgiveness from another, leading to guilt
and despair. In a spiritual context, the end of life may trig-
ger a faith crisis with respect to an unjust God who would
permit death to occur or to ineffectual prayer for healing,
leading to decreased trust or hope in God’s help.
Expectations of Increase
On the other hand, from the psychological perspectives of
Erikson (1968) and Jung (1933), one would make quite op-
posite inferences about religiousness at the end of life. Erik-
son’s influential eight stages of human psychosocial devel-

S328
IDLER ET AL.
as prayer (Ellison & Taylor, 1996; Harrison, Koenig, Hays,
Eme-Akwari, & Pargament, 2001; Pargament et al., 1990).
Hence this perspective led us to expect increases in both
religious behavior and belief in those meeting the height-
ened adaptational challenges of old age.
Thus with conflicting theory failing to guide us ade-
quately, we undertook an exploratory analysis of trends in
religious behavior and belief in the period immediately pre-
ceding death. We examined cross-sectional differences be-
tween respondents who were and were not in their last 12
months of life for several indicators of religiousness. We
also looked at change within respondents over time for three
waves of data. Because the health status of respondents
might have affected either their capacity to attend services
or their religious feelings, we adjusted our analyses for sev-
eral measures of health status. Analyses were also adjusted
for age, sex, race, education, marital status, and religious
preference.
Figure 1. New Haven sample, 1982–95, deaths occurring in the
year following interviews that included religion items. Shaded areas
of survival curve represent deaths in 12 months immediately follow-
ing face-to-face interviews. n ꢀ 499. Data are from the Established
Populations for Epidemiologic Studies of the Elderly.
METHODS
Sample
Data for the study came from the New Haven site of the Es-
tablished Populations for Epidemiologic Studies of the El-
derly (EPESE). Data collection procedures have been de-
scribed elsewhere (Cornoni-Huntley, Brock, Ostfeld, Taylor,
& Wallace, 1986). The stratified, weighted sample was
representative of the elderly population of metropolitan
New Haven, Connecticut (N ꢀ 2,812). Participants were in-
terviewed annually in their homes or by telephone from
1982 to 1989 and once again in 1994. Mortality surveillance
for the sample was continuous through 1996, and loss to fol-
low-up was less than 1%. Virtually all deaths were con-
firmed by death certificate.
Response rates at baseline and at annual follow-up inter-
views were quite high for a community study. This is criti-
cal to the feasibility of this project, because selective initial
response and attrition could be expected to have a dispro-
portionate impact on respondents in their last year of life. At
baseline, 82% of sampled respondents completed inter-
views, and response rates for annual reinterviews averaged
94–96% of the surviving sample. When a respondent was
too ill to provide usable information, a close family member
or care provider was asked to provide basic information.
Proxies did not respond to items that called for the respon-
dent’s judgment or attitude. Deaths began occurring in the
sample shortly after interviewing began. Figure 1 shows the
time period of the follow-up. Of the 2,812 respondents who
began the study in 1982, just 939 survived for 12 months
following their final interview in 1994.
gious involvement. Intervals from interview to death were
as short as 2 days, and the distribution of deaths during the
year was evenly spread: 22 of the year’s deaths (16.8%) oc-
curred in the first 90 days following the 1982 interview; 31
deaths (22.3%) in the 90 days following the 1985 interview;
29 deaths (25.5%) in the 90 days following the 1988 inter-
view; and 24 deaths (25.0%) in the 90 days following the
1994 interview. Scores for change from the previous inter-
view were available for the 350 last-year-of-life respon-
dents with interviews in 1985, 1988, or 1994 and who had
also been interviewed 3 years previously, or 6 years for the
last pair of interviews.
Measures
We analyzed three dependent variables measuring reli-
giousness. Responses to the question about attendance at re-
ligious services (“About how often do you go to religious
meetings or services?”) were recoded from six ordinal cate-
gories to number of times per year they represented. Those
who never attended were coded 0 (times per year); those
who attended once or twice a year were coded 1; those who
attended every few months were coded 4; those who at-
tended once or twice a month were coded 18; those who at-
tended once a week were coded 52; and those who attended
more than once a week were coded 104. Regular twice-
weekly attendance at services is common for some churches
in New Haven (Griffith, English, & Mayfield, 1980). A sec-
ond measure emphasized the subjective religious identity of
the respondent (“Aside from attendance at religious ser-
vices, do you consider yourself to be deeply religious, fairly
religious, only slightly religious, not at all religious, or
against religion?”). This item was coded so that the last two
categories were combined, and the high score was for
deeply religious (4). Finally, we assessed feelings of benefit
from religion (“How much is religion a source of strength
Questions about religiousness were asked only in the
longer, face-to-face interviews administered in 1982, 1985,
1988, and 1994. To calculate time to death from these inter-
views, we subtracted the date of the respondent’s 1982,
1985, 1988, or 1994 interview from the date of their death.
Intervals less than 183 days made up the 6 months or less
survival group, and intervals of 183–365 days made up the
6–12 months survival group. Our last-year-of-life group
thus consisted of 499 respondents who died within 365 days
following an interview in which they were asked about reli-

RELIGIOUSNESS IN THE LAST YEAR OF LIFE
S329
and comfort to you?”). This item was coded 1 (none) to 3 (a
great deal). All items were analyzed separately. Proxies
were asked the item for attendance at services, but not the
subjective religiousness items.
the previous interview but before the next one. For our lon-
gitudinal analyses, we created difference scores for each re-
ligion item by subtracting the respondent’s score in 1982,
1985, or 1988 from his or her score in the next round of in-
terviews (1985, 1988, 1994) and treated the differences as
dependent variables in covariance models adjusted for the
same sociodemographic and health status measures as in the
cross-sectional analyses.
Our measures of health status included functional ability,
prescription medications, hospitalizations, nursing home
stay, and bed disability. All health status measures were
asked identically at each of the four interviews, and for each
model we used the concurrent health status variables. Func-
tional ability was assessed with a 15-item scale with scores
that ranged from 0 to 150 (ꢁ ꢀ .87 for the 1982 scale; for
details, see Idler & Kasl, 1997a). Prescription medications
were observed in the home by the interviewer; we used
the number of different medications the respondent had taken
in the past 2 weeks. Respondents were asked if they had
been hospitalized during the previous year, if they had ever
been a patient in a nursing home, and if they had spent more
than 1 week in bed during the previous 3 months; each item
was coded 1 for yes and 0 for no. The respondent’s sex and
age in years were included in all analyses. Finally, we ad-
justed for religious preference and race jointly by creating
dummy variables for White Protestants, Black Protestants,
Jews, and others/none and comparing them with the refer-
ence category of Roman Catholics.
RESULTS
Our first concern was that, even if our near-death respon-
dents had agreed to interviews with very high response
rates, there could still be substantial amounts of missing
data within the interviews that would affect the analysis. A
second concern was that the last-year-of-life group would
be very different demographically from the total sample.
These issues are addressed in Table 1.
Table 1 compares baseline measures for the 499 last-
year-of-life respondents with the total 2,812 who began the
study. The table shows that the study was feasible because
the amount of missing data was negligible. With the single
example of the interviewer-observed prescription medica-
tion measure, the amount of missing data was extremely
small for all items, including religion items. Although the
rates of missing data were slightly larger in the last-year-of-
life sample, the levels were such that there should be no
concern. Table 1 also shows that the last-year-of-life sample
was quite similar to the total sample from which it came.
The last-year-of-life group had more risk factors for mortal-
ity, including a higher proportion of men, a higher average
age, more disability, more prescription medications, more
hospitalizations, more nursing home stays, and more bed
disability than the sample as a whole, but none of these dif-
ferences was large. The religious affiliations of the last-
year-of-life sample were similar to the total sample, with
slightly fewer Protestants and slightly more Jews. No statis-
tical tests were performed because the groups were not mu-
tually exclusive.
Analysis
We began by comparing the last-year-of-life sample with
the total baseline sample with respect to missing data. For
our cross-sectional analyses, we estimated analysis of cova-
riance models with each of the three religious involvement
items in 4 years of follow-up as the dependent variable and
religious preference/race and sociodemographic factors as
covariates; in a second step we adjusted for current health
status. We treated time to death following each of the four
interviews as a set of two dummy variables, one for 0–6
months and one for 6–12 months, with 12-month survival as
the reference group. Thus the two end-of-life groups were
completely different for each year of analysis, whereas the
comparison group of survivors remained somewhat the
same, losing only those who died 12 months or longer after
We present the cross-sectional data on patterns of reli-
giousness in Tables 2 and 3. Table 2 shows each of the three
Table 1. 1982 Characteristics of Total Sample and Last Year of Life (LYOL) Sample, New Haven EPESE
LYOL Samplea (n ꢀ 499) Total Sampleb (N ꢀ 2,812)
% Missing Data % Missing Data
Variable
%
%
Male
Roman Catholic
White Protestant
Black Protestant
Jewish
Other/none
Age (m)
Hospitalization in past year
Ever patient in nursing home
1 or more weeks in bed last 3 months
No. of prescription medications (m)
Functional disability (m)
49.5
51.9
10.8
14.8
14.6
7.8
76.6
28.1
6.7
0.0
0.0
0.0
0.0
0.0
0.0
0.0
1.6
1.4
1.6
3.0
1.2
41.5
52.0
11.8
16.0
13.3
7.0
74.5
19.6
3.9
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.7
0.7
0.7
4.5
1.1
10.0
3.3
27.3
6.2
2.7
18.8
Note: EPESE ꢀ Established Populations for Epidemiologic Studies of the Elderly.
aCombines all those who died within 1 year of 1982, 1985, 1988, and 1994 interviews.
bEntire sample, including LYOL sample.

S330
IDLER ET AL.
Table 2. Attendance, Religiousness, and Strength From Religion Regressed on YHAP Respondents’ Distance from Death, Demographic,
and Health Status Variables, New Haven EPESE Data, 1982
Attendance
Religiousness
Strength
Variable
Model 1
Model 2
Model 1
Model 2
Model 1
Model 2
Survived less than 6 months following 1982 interviewa
.38b
ꢂ7.03*
ꢂ5.13***
ꢂ.02
.29
ꢂ1.31
ꢂ12.08***
1.89
ꢂ13.54***
ꢂ10.31***
5.28
ꢂ2.40
ꢂ6.82***
.13
.15
ꢂ.42
ꢂ11.95***
3.93*
ꢂ12.74***
ꢂ9.74***
ꢂ.27***
ꢂ.02
.18
ꢂ.13
.16
ꢂ.14
.03
ꢂ.10
ꢂ.22***
.01**
ꢂ.01
.02
ꢂ.12
ꢂ.23***
.01**
ꢂ.01
Survived 6–12 months following 1982 interviewa
Male
Age
Education
ꢂ.27***
.01***
ꢂ.02***
ꢂ.08*
ꢂ.29***
.12**
ꢂ.27***
.01***
ꢂ.02***
ꢂ.08*
ꢂ.31***
.11**
ꢂ.52***
ꢂ.31***
ꢂ.00
Married
ꢂ.04
ꢂ.04
White Protestantc
ꢂ.20***
.13***
ꢂ.46***
ꢂ.25***
ꢂ.21***
.12***
ꢂ.46***
ꢂ.25***
ꢂ.00
Black Protestantc
Jewishc
ꢂ.50***
ꢂ.31***
Other/nonec
Functional disability for current year
No. of prescription medications for current year
.01*
.01*
Hospitalization in past year
ꢂ.97
.02
.01
Nursing home stay in previous year
1 or more weeks in bed last 3 months
ꢂ6.18*
ꢂ.21
ꢂ.00
.09
ꢂ.06
.04
Intercept
R2
n
32.39
.05
2,714
27.45
.10
2,587
2.62
.13
2,684
2.40
.13
2,563
2.56
.14
2,663
2.44
.14
2,540
Notes: EPESE ꢀ Established Populations for Epidemiologic Studies of the Elderly; YHAP ꢀ Yale Health and Aging Project.
aReference category: survived 12 months or longer.
bUnstandardized parameter estimates from analysis of covariance models.
cReference category: Roman Catholic.
*p ꢃ .05; **p ꢃ .01; ***p ꢃ .001.
measures of religiousness regressed on time to death and a
set of sociodemographic and then health status variables, for
1982 only. Table 3 displays the least squares means cal-
culated for each religiousness measure for each survival
group in all years, with post hoc tests of significance. This
approach shows two different but equally important aspects
of the analysis: the relative importance of the last year of
life as a determinant of religiousness (in Table 2), and the
absolute levels of religious involvement of respondents in
the three time-to-death groups (in Table 3).
Table 3. Least Squares Means and Tests of Differences for Religion
Variables, by Distance From Death, for Each Year, New Haven
EPESE Data
Table 2 presents analysis for 1982 only; similar analyses
were conducted for 1985, 1988, and 1994, and the findings
were parallel in all important respects. Two models are pre-
sented for each of the three dependent variables; in the first
step, time to death was entered along with sociodemo-
graphic factors. In the second step, health status covariates
were added. This analytic approach allowed us to see the
impact of the last year of life on religiousness, both before
and after current health status was taken into account. The
first column shows that respondents in their last 6 months of
life did not attend religious services more or less than
those who survived more than 12 months, and those who
lived only 6–12 months attended significantly less often.
Men attended less often than women, and White Protes-
tants, Jews, and those of other religions attended less often
than Roman Catholics. These factors explained only about
5% of the variance in religious attendance. In Model 2 we
added current health status, which doubled the variance ex-
plained, to 10%. If poor health status reduced the ability of
respondents to attend religious services, then the apparently
lower levels of those at the end of life should be adjusted
upward by the inclusion of these variables. This was in fact
what happened; the lower levels of attendance for those in
their last 6–12 months of life were reduced to being insig-
nificantly different from those who survived, and those in
their last 6 months of life showed attendance levels higher
(but not significantly so) than those of the survivors. As for
Variable
Distance from Death
1982
1985
1988
1994
Attendancea
ꢃ6 months
6–12 months
ꢄ12 monthsb
n
25.1
17.4
19.8
19.6
25.2
26.7
24.3
30.1
27.1
11.9*
24.5
23.4
2,587
2,037
1,541
837
Religiousnessc
ꢃ6 months
6–12 months
ꢄ12 monthsb
n
3.2
2.9
3.1
3.1
3.1
3.2
3.3*
3.1
3.0
2.9
2.9
3.0
2,563
1,907
1,392
702
Strengthd
ꢃ6 months
6–12 months
ꢄ12 monthsb
n
2.6
2.4
2.5
2.6
2.5
2.5
2.5
2.5
2.5
2.3
2.4
2.4
2,540
1,903
1,379
694
Notes: Least squares means estimated from models adjusted for sex, age, ed-
ucation, marital status, functional disability, prescription medications, hospital-
ization, nursing home stay, extended bed disability, and religious preference/race.
EPESE ꢀ Established Populations for Epidemiologic Studies of the Elderly.
aTimes per year attended.
bReference category: survived 12 months or longer following the interview,
for each year.
c1 ꢀ not at all religious, 4 ꢀ deeply religious.
d1 ꢀ no comfort, 3 ꢀ a great deal.
*p ꢃ .05.

RELIGIOUSNESS IN THE LAST YEAR OF LIFE
S331
the health status variables themselves, poor functional abil-
ity and previous nursing home stays both acted to signifi-
cantly reduce attendance.
sons in their last 6 months of life in the last year of follow-
up, to a high of 30, for persons in their last 6–12 months of
life in 1988; where there was a significant difference, it was
in the direction of lower attendance for the soonest to be de-
ceased. Average levels of how deeply religious respondents
felt were at or above the equivalent of “fairly” (3 on a
4-point scale); where there was a difference here, it was in
the direction of deeper feelings of religiousness for the
short-term survivors. Average levels of strength and com-
fort received from religion were also high, mostly 2.5 and
higher on a 3-point scale, with no differences to speak of.
The next set of analyses moved from the cross-sectional
to the longitudinal. We examined group differences in
changes from the previous round of interviews, adjusted for
age, sex, marital status, education, and the health status of
the respondents at the later time. The bar graphs show an in-
crease or decrease in religiousness as bars above or below
the center line, which represents no change. Figure 2 shows
a slight increase in attendance from 1982 to 1985 for the
6–12-month and 12-month survivor groups, but it shows a
significant drop of six times per year for the 0–6-month survi-
vor group. In 1994 as well there was a significant drop in at-
tendance from previous levels for the 0–6-month group,
compared with the 12-month survivors. Recall from Table 3
that the absolute level of attendance in this population was
very high.
Figures 3 and 4, showing changes in subjective reli-
giousness, are very similar to each other. The close proxim-
ity of the bars to the center line shows the essential stability
of these measures. Respondents’ reports of religious feel-
ings showed almost no change, but there was a statistically
significant difference in the direction of greater religious-
ness for the near-death group in 1988 when they were com-
pared with the 12-month survivors, who showed a slight
decline. There were also significant differences for the
6-month group in Figure 4; those with less than 6 months to
live in 1985 said they were getting slightly more strength
and comfort from religion, and when compared with the
tiny decline for the 12-month survivors, the difference was
statistically significant. In all other groups change was
barely perceptible, showing extraordinary stability over the
years both for those who were in the last year of life and
those who were not. Again, the absolute level of this stabil-
ity was very high.
The second pair of columns has a much simpler story to
tell. With respect to how religious they believed they were,
it made no difference whether or not the respondents were
in their last year of life or not. Many other factors did differ-
entiate within the sample: Men believed they were less reli-
gious than women; older people believed they were more
religious than younger elderly people; those with more edu-
cation and those who were married reported they were less
religious than those who were less educated and unmarried;
and White Protestants, Jews, and those of other religions
felt less religious than Roman Catholics, whereas Black
Protestants reported feeling more religious. Interestingly,
none of these coefficients changed in any meaningful way
when the health status variables were introduced in Model
2. None of the health status measures were significantly cor-
related with feelings of religiousness, with the exception of
the number of prescription medications—the more medica-
tions, the higher the feelings of religiousness.
The final pair of models was rather similar to the second
pair. Respondents who were in their last year of life did not
differ in their feelings of strength and comfort from religion
from those who survived. Other sociodemographic differ-
ences were also similar: Men; younger respondents; and
White Protestants, Jews, and those of other religions felt
less strength and comfort than women, more elderly respon-
dents, Roman Catholics, and Black Protestants. The same
indicator of poorer health status (number of prescription
medications) was also associated with greater comfort from
religion. Thus, of the numerous influences on feelings of re-
ligiousness in this cross-sectional sample of elderly people,
whether or not one was in his or her last year of life ap-
peared rather unimportant, and it mattered little whether or
not health status was considered. These models explained
13–14% of the variance in religious feelings.
Table 3 examines these variables more closely, again
cross-sectionally, in each of the 4 years. For each year, ad-
justed least squares means were calculated from the multi-
variate models presented in Table 2 (and replicated for the
other years in analyses not presented). Post hoc tests of dif-
ferences were run, comparing the means for each of the two
last-year-of-life groups to the 12-month survivors. Briefly,
there were only two significant post hoc differences among
the 24 tests, a number only slightly above what one would
expect by chance. Those who died in the next 6 months at-
tended services significantly less often (in 1994) and were
more deeply religious (in 1988) than who lived 12 months
or more. Elsewhere, although there were no statistically
significant differences, persons in their last 6 months, and
last 6–12 months of life were sometimes more frequent at-
tenders, sometimes less, when compared with those who
survived 12 months or longer. The two measures of subjec-
tive religious feelings showed only very small differences,
again in both directions. The most important feature of the
table, however, is not the differences or lack of them, but
the quite high average levels of involvement for all of the
groups. Average attendance at religious services ranged
from a low of 12 times per year, or once a month, for per-
Figure 2. Change in attendance from previous interview, 1985,
1988, and 1994. Means were calculated from models adjusted for all
covariates as in Table 2. N ꢀ 2,812. Data are from the New Haven site
of the Established Populations for Epidemiologic Studies of the El-
derly. *p ꢃ .05 for comparison of means with 12 or more months.

S332
IDLER ET AL.
months of life we did some additional analyses (not shown),
testing differences between the whole group in the last 12
months of life and those who survived 12 months or longer.
We found that there were no differences at all between these
two groups, in attendance or religious feelings. For this rea-
son, we retain the focus on the last 6 months of life.
Also, the apparently lower levels of attendance in the
6–12-month group, when compared with the 12-month sur-
vivors, are almost entirely attributable to the poorer health
of these respondents, because the significantly lower levels
of attendance in the last-year-of-life group disappear when
the analysis is adjusted for health status. The pattern of
lower attendance is consistent with the multidimensional
disengagement theory, in the sense that attendance would be
the marker of religiousness as a social role. However, this
decline in social engagement comes extremely late in the
life course, and is clearly incomplete, if average levels of at-
tendance remain as high as they do in the period immedi-
ately before death.
The stability/slight increase in the subjective measures of
religiousness also supports the multidimensional disengage-
ment perspective, because there is no decline here at all.
Given the usual high levels of stability in attitudinal mea-
sures such as these, any change at all, and particularly an in-
crease when the base levels are already so high, is worth
noting. It is also noteworthy that health status has no impact
on subjective feelings of religiousness. The possibility that
terminal decline might affect religious and spiritual beliefs
also has no support, because our more abstract, cognitive in-
dicators of religiousness as identity or source of solace
show no decline at all.
It appears that the multidimensional disengagement the-
ory accounts for the patterns we see in our sample in that the
behavioral and subjective measures of religious involve-
ment behave differently. We also argue that the religious
coping perspective contributes additional insight, in that the
findings in Table 2 link poor health to higher levels of sub-
jective religious feelings. The items themselves suggest that
they are coping responses; respondents with higher numbers
of prescription medications believe that they get more
strength and comfort from religion and rate themselves as
more religious than those taking fewer medications.
However, the approximate fit of our data with these theo-
retical schemes is not the most important finding of the
study. We would rather emphasize the descriptive nature of
the study, as a first look at the natural history of several
forms of religious involvement in a large sample of elderly
people approaching the end of life. It is far more important
that we underscore that the lowest average attendance for
any group studied is still quite high. For those with less than
6 months to live at the time of the final follow-up, the aver-
age attendance is still 12 times per year, or once a month.
Thus we argue that it is the absolute, and not the relative,
levels of religious observance and belief that are notable
findings in this study.
Figure 3. Change in religiousness from previous interview, 1985,
1988, and 1994. Means were calculated from models adjusted for all
covariates as in Table 2. N ꢀ 2,812. Data are from the New Haven site
of the Established Populations for Epidemiologic Studies of the El-
derly. *p ꢃ .05 for comparison of means with 12 or more months.
The trends in religiousness in the period before death in
this sample were thus predictable extensions of the cross-
sectional patterns. There were distinct declines in atten-
dance from the respondents’ own reports of attendance at
the earlier interview. For the group that died in the 6 months
following their 1985 interview, for example, the average
level of attendance in 1985 was 19.6 times per year (Table
3). This is still an average level of attendance that well ex-
ceeds once per month, even in a group in the last few
months of life. The more subjective, spiritual measures
showed some evidence of an increase in religiousness at the
end of life. These were certainly not large increases, nor
were the overall differences between the near-deceased and
the 12-month survivors large ones, but there may have been
a ceiling effect operating because the levels for all groups
were already so high.
DISCUSSION
In summary, our findings provide mixed support for the the-
oretical schema we presented at the beginning of the article.
The findings do show evidence of declining attendance in
the period immediately before death; this decline is evident
for the last half year of life but not for the 6–12 months pre-
ceding it. This pattern would be consistent with the interpre-
tation that the respondents in their last 6 months of life are
experiencing rapid declines in health and functioning. To
show that this decline is limited to those in their last 6
Figure 4. Change from previous interview in receiving strength
and comfort from religion, 1985, 1988, and 1994. Means were calcu-
lated from models adjusted for all covariates as in Table 2. N ꢀ 2,812.
Data are from the New Haven site of the Established Populations for
Epidemiologic Studies of the Elderly. *p ꢃ .05 for comparison of
means with 12 or more months.
These findings cannot be considered definitive; both the
data and the analyses conducted here have limitations. This
is a regional sample, and its particular ethnic and religious
makeup may make the findings not generalizable to other
elderly populations. The elderly population of New Haven

RELIGIOUSNESS IN THE LAST YEAR OF LIFE
S333
is quite diverse in that it contains solid proportions of Chris-
tians, both Protestants and Roman Catholics, and Jews, and
about half of the Protestants are African American. How-
ever, the sample overrepresents Roman Catholics and Jews
and underrepresents Protestants, when compared with na-
tional figures. Moreover, it has only a tiny Hispanic (mostly
Roman Catholic) representation and very few Muslims (in-
cluded in “other religion”). As in many old northeastern cit-
ies, traditional neighborhoods divided by race, religion, and
ethnicity have remained stable over generations. Churches
and temples persist as the cultural centers of these geo-
graphical territories, providing services, support, and a
sense of identity to congregation members who also know
each other as family members and neighbors.
Perhaps a second limitation of this study is the uncer-
tainty of the meaning of being in the last year of life.
“Awareness of dying” has long been a topic of interest for
social scientists studying populations of the terminally ill
(Glaser & Strauss, 1965; Seale, 1991). These studies have
concluded that there is a considerable range of awareness of
how close to death one is, even among hospitalized, termi-
nally ill patients with untreatable cancers. In a community
sample such as ours, we must assume that there would be
considerably less awareness on average than in a clinical
sample. If many or even most study participants are un-
aware of how much time they have left, should we expect to
see any differences in behavior or feelings at all?
At the same time, this uncertainty may be seen as a
strength of the study. The potential for learning about the
last year of life in a wide range of people is in itself an excit-
ing development. As Lentzner, Pamuk, Rhodenhiser, Roth-
enberg, and Powell-Griner (1992) and Lawton, Moss, and
Glicksman (1990) pointed out, most of the data on the qual-
ity of life in the last year of life comes from studies of indi-
viduals who have been identified as terminally ill, such as
respondents to the National Hospice Study (Morris, Suissa,
Sherwood, Wright, & Greer, 1986) or the Study to Under-
stand Prognoses and Preferences for Outcomes and Risks of
Treatments (SUPPORT; Knaus & SUPPORT Investigators,
1995), or from hospital or nursing home patients; the em-
phasis has been on the quality of life as a function of care
for the dying. Seale and Cartwright (1994), Lentzner and
colleagues (1992), Lawton and colleagues (1990), and more
recently Liao, McGee, Cao, and Cooper (2000) have at-
tempted to move beyond studies of patients, to study the last
year of life of respondents living in the community. How-
ever, these investigators identified individuals for their stud-
ies by sampling death certificates and have thus been lim-
ited by biases inherent in retrospective methods of data
collection and the use of proxy respondents. Authors of
these studies were aware of and directly addressed the prob-
lems of bias in proxy response, but they noted that it is ex-
tremely difficult to identify elderly people who are in their
last year of life prospectively. The only feasible way of do-
ing that is by having a large sample recruited with a high re-
sponse rate, followed for a long period with little attrition
other than mortality, and interviewed annually, so that an
interview takes place in virtually every last year of life.
This makes the design and execution of the EPESE stud-
ies ideal for studies of the quality of the last year of life of
elderly people living in the community. Such studies should
become even more important in the future, if the compres-
sion of morbidity and disability into the very last period of
life continues (Fries, 1980; Manton, Corder, & Stallard,
1997) and successful aging increasingly means dying of old
age rather than disease (Nuland, 1993; Rowe & Kahn,
1998). A recent Institute of Medicine (1997) report made
the point that clinical care of the dying has proceeded with-
out population data on norms or outcomes. Our study exem-
plifies the promise of population data for study of the entire
population of elderly people, not just the minority who are
being treated in hospitals, nursing homes, or hospices. We
see this study as an effort to describe the social life condi-
tions of a representative sample of the entire population of
elderly persons in this community; it is a public health per-
spective rather than a clinical one.
The significance of religion at the end of life is often ac-
cepted as a matter of faith. The institution of the modern
hospice, providing specialized care for the dying, has tradi-
tionally emphasized spiritual aspects of care. The founder of
the hospice movement, Dame Ciceley Saunders, explicitly
conceived of the hospice as a quasi-religious institution, re-
garding the spiritual care of the dying as central to the mis-
sion of the institution (du Boulay, 1984). The current ver-
sion of the primary text in palliative care, the Oxford
Manual of Palliative Care, continued this emphasis (Doyle,
Hanks, & MacDonald, 1998). The often contentious debates
about the place of religion and spirituality in modern medi-
cal care (Sloan, Bagiella, & Powell, 1999) are subdued when
the subject becomes end-of-life care. The influence of tradi-
tional religious beliefs on the decisions of patients and fam-
ilies about subjects such as advance directives or the impor-
tance of last rites is usually acknowledged and respected,
even in secular hospitals (Grodin, 1993). Our findings sup-
port the wisdom of such respect. Despite a perhaps predict-
able occasional decline in attendance at public worship ser-
vices during the last few months of their lives, our
respondents demonstrate no decline, and even some in-
crease in subjective feelings and beliefs about their faith.
Evidence from this community study underscores the need
to keep these resources available for those who may no
longer be able to seek them out for themselves.
Acknowledgments
The first author was supported by Grant R01 AG16750 from the Na-
tional Institute on Aging. The third author was supported by funding from
the Fetzer Foundation, Kalamazoo, MI, and from the National Institutes of
Health, National Institute on Aging, Grants K01AG00856-02 and N01-AG-
1-2102 in support of the Established Populations for Epidemiologic Studies
of the Elderly (Duke).
We thank the staff of the Yale Health and Aging Project, the New Haven
site of the Established Populations for Epidemiologic Studies of the El-
derly, Diane Davis, Julie McLaughlin, Allen Buurma, and DuWayne Battle.
Address correspondence to Ellen Idler, Institute for Health, Health Care
Policy, and Aging Research, Rutgers University, 30 College Ave., New
Brunswick, NJ 08903. E-mail: idler@rci.rutgers.edu
References
Berg, S. (1996). Aging, behavior, and decline. In J. Birren & K. W. Schaie
(Eds.), Handbook of the psychology of aging (4th ed., pp. 323–337).
New York: Academic Press.

S334
IDLER ET AL.
Bevins, M., & Cole, T. (2000). Ethics and spirituality: Strangers at the end
of life? In M. P. Lawton (Ed.), Annual review of gerontology and geri-
atrics (Vol. 20, pp. 16–38). New York: Springer.
Blazer, D., & Palmore, E. (1976). Religion and aging in a longitudinal
panel. The Gerontologist, 16, 82–85.
Knaus, W., & SUPPORT Investigators. (1995). A controlled trial to im-
prove care for seriously ill hospitalized patients: The Study to
Understand Prognoses and Preferences for Outcomes and Risks of
Treatments (SUPPORT). Journal of the American Medical Associa-
tion, 274, 1591–1598.
Bradley, E. H, Fried, T. R., Kasl, S. V., & Idler, E. (2000). Quality of life
trajectories in the end of life. In M. P. Lawton (Ed.), Annual review of
gerontology and geriatrics, (Vol. 20, pp. 64–96). New York: Springer.
Cornoni-Huntley, J., Brock, D. B., Ostfeld, A. M., Taylor, J. O., & Wal-
lace, R. B. (Eds.). (1986). Established Populations for Epidemiologic
Studies of the Elderly: Resource data book (NIH Publ. No. 86-2443).
Bethesda, MD: National Institute on Aging.
Cumming, E., & Henry, W. E. (1961). Growing old: The process of disen-
gagement. New York: Basic Books.
Doyle, D., Hanks, G. W. C. & MacDonald, N. (1998). Oxford textbook of
palliative medicine (2nd ed.). Oxford, England: Oxford University
Press.
du Boulay, S. (1984). Ciceley Saunders: Founder of the modern hospice
movement. London: Hodder and Stoughton.
Ellison, C. G., & Taylor, R. J. (1996). Turning to prayer: Social and situa-
tional antecedents of religious coping among African Americans.
Review of Religious Research, 38, 111–131
Koenig, H. (1995). Religion and health in later life. In M. A. Kimble &
S. H. McFadden (Eds.), Aging, spirituality, and religion: A handbook
(pp. 9–29). Minneapolis, MN: Fortress Press.
Lawton, M. P., Moss, M., & Glicksman, A. (1990). The quality of the last
year of life of older persons. Milbank Quarterly, 68, 1–28.
Lentzner, H. R., Pamuk, E. R., Rhodenhiser, E. P., Rothenberg, R., & Pow-
ell-Griner, E. (1992). The quality of life in the last year before death.
American Journal of Public Health, 82, 1093–1098.
Levin, J., & Markides, K. (1985). Religion and health in Mexican Ameri-
cans. Journal of Religion and Health, 24, 60–69.
Liao, Y., McGee, D., Cao, G., & Cooper, R. (2000). Quality of the last year
of life of older adults: 1986 vs 1993. Journal of the American Medical
Association, 283, 512–518.
Maier, H., & Smith, J. (1996, November). Psychological functioning and
distance from death: Initial results from the Berlin Aging Study. Paper
presented at the 49th Annual Scientific meeting of The Gerontological
Society of America, Washington, DC.
Erikson, E. (1968). Identity, youth, and crisis. New York: Norton.
Erikson, E. H., Erikson, J. M., & Kivnick, H. Q. (1986). Vital involvement
in old age. New York: Norton.
Manton, K. G., Corder, L., & Stallard E. (1997). Chronic disability trends
in elderly United States populations, 1982–1994. Proceedings of the
National Academy of Science, 94, 2593–2598.
Fowler, J. W. (1981). Stages of faith. San Francisco: Harper.
Fries, J. (1980). Aging, natural death and the compression of morbidity.
New England Journal of Medicine, 303, 130–135.
Glaser, W., & Strauss, M. R. (1965). Awareness of dying. Chicago: Aldine.
Griffith, E. E. H., English, T., & Mayfield, V. (1980). Possession, prayer,
and testimony: Therapeutic aspects of the Wednesday night meeting in
a Black church. Psychiatry, 43, 120–128.
Mindel, C., & Vaughan, C. E. (1978). A multidimensional approach to reli-
giosity and disengagement. Journal of Gerontology, 33, 103–108.
Morris, J. N., Suissa, S., Sherwood, S., Wright, S. M., & Greer, D. (1986).
Last days: A study of the quality of life of terminally ill cancer patients.
Journal of Chronic Diseases, 39, 47–62.
Nuland, S. (1993). How we die: Reflections on life’s final chapter. New
York: Vintage.
Grodin, M. (1993). Religious advance directives: The convergence of law,
religion, medicine, and public health. American Journal of Public
Health, 83, 899–903.
Harrison, M. O., Koenig, H. G., Hays, J. C., Eme-Akwari, A. G., & Parga-
ment, K. (2001). The epidemiology of religious coping: A review of
recent literature. International Review of Psychiatry, 13, 86–93.
Hays, J., Landerman, L., Blazer, D., Koenig, H., Carroll, J., & Musick, M.
(1998). Aging, health and the “electronic church.” Journal of Aging
and Health, 10, 458–482.
Idler, E., & Kasl, S. V. (1997a). Religion among disabled and nondisabled
persons: I. Cross-sectional patterns in health practices, social activities,
and well-being. Journal of Gerontology: Social Sciences, 52B, S294–
S305.
Idler, E., & Kasl, S. V. (1997b). Religion among disabled and nondisabled
persons: II. Attendance at religious services as a predictor of the course
of disability. Journal of Gerontology: Social Sciences, 52B, S306–
S316.
Institute of Medicine. (1997). Approaching death: Improving care at the
end of life. Washington, DC: Author.
Jung, C. (1933). Modern man in search of a soul (W. S. Dell & C. F.
Baynes, Trans.). New York: Harcourt, Brace.
Kimble, M. A., McFadden, S., Ellor, J., & Seeber, J. (Eds.). (1995). Aging,
spirituality, and religion: A handbook. Minneapolis, MN: Fortress
Press.
Pargament, K. (1997). The psychology of religion and coping: Theory, re-
search, and practice. New York: Guilford.
Pargament, K. I., Ensing, D. S., Falgout, K., Olsen, H., Reilly, B., Van Hai-
tsma, K., & Warren, R. (1990). God help me: I. Religious coping
efforts as predictors of the outcomes to significant negative life events.
American Journal of Community Psychology, 18, 793–824.
Pargament, K., Van Haitsma, K., & Ensing, D. (1995). Religion and cop-
ing. In M. A. Kimble & S. H. McFadden (Eds.), Aging, spirituality, and
religion: A handbook (pp. 47–67). Minneapolis, MN: Fortress Press.
Ploch, D., & Hastings, D. (1994). Graphic presentations of church atten-
dance using general social survey data. Journal for the Scientific Study
of Religion, 33, 16–33.
Rowe, J., & Kahn, R. (1998). Successful aging. New York: Pantheon.
Seale, C. (1991). Communication and awareness about death: A study of a
random sample of dying people. Social Science and Medicine, 32, 943–
952.
Seale, C., & Cartwright, A. (1994). The year before death. Aldershot, En-
gland: Avebury.
Siegler, I. (1975). The terminal drop hypothesis: Fact or artifact. Experi-
mental Aging Research, 1, 169–185.
Sloan, R., Bagiella, P., & Powell, T. (1999). Religion, spirituality, and
medicine. Lancet, 353, 664–667.
Kleemeier, R. (1962). Intellectual changes in the senium. Proceedings of
the American Statistical Association, 1, 290–295.
Received July 5, 2000
Accepted April 18, 2001