Abstract
The aim of this study was to examine the association of religiosity and death attitudes with self-reported advance care planning (ACP) in chronically ill older adults. Survey data were collected in person for a sample of 157 chronically ill older adults drawn from primary care clinics in North Carolina. Logistic regression was used to examine associations of religiosity and death attitudes in the likelihood of engagement in three ACP outcomes: (a) ACP discussions with the doctor, (b) ACP discussions with family, and (c) the completion of a living will. Greater reported religiosity (b = 1.67, p < .01) was significantly associated with reported ACP discussions with the doctor (R2 = .29, model significance p < .01). Less fear of death was significantly associated (b = −0.41, p < .01) with self-reported completion of a living will (R2 = .21, model significance p < .01). Religiosity and fears of death should be considered in future ACP studies.
Advance care planning (ACP) has been advocated as a way to improve end-of-life (EOL) care of seriously ill patients (Field & Cassel, 1997; Teno, Gruneir, Schwartz, Nanda, & Wetle, 2007). In the United States, an advance directive (e.g., a living will, health care power of attorney) is a central component of ACP (Lo & Steinbrook, 2004) and provides patients with the ability to identify their care preferences, even if they lack decision making capacity. ACP is a more multifaceted process than completion of an advance directive, one that includes discussions in which patients, families, doctors, and other health care providers explore goals of care in the context of current and potential health declines, discuss treatment options, and document patients’ wishes (Morrison & Meier, 2004). For example, communication with health care professionals, whether initiated by patients or health care professionals, and communication with family members are important components of ACP.
Despite broad support given to the early assessment of patient care preferences, and the emphasis of documenting these preferences as advance directives, intervention studies of ACP have had limited success in terms of implementing patient care preferences and reducing life-sustaining treatments (Fried, Tinetti, & Iannone, 2010; Ko & Lee, 2010; Schickedanz et al., 2009; Schneiderman, Kronick, Kaplan, Anderson, & Langer, 1992; Sudore & Fried, 2010; Sudore et al., 2008; Sudore, Schillinger, Knight, & Fried, 2010; SUPPORT Principal Investigators, 1995). There are several reasons why ACP interventions have had limited success (Hare & Nelson, 1991; Heiman, Bates, Fairchild, Shaykevich, & Lehmann, 2004; High, 1993; Sachs, Stocking, & Miles, 1992; Sulmasy, Terry, Faden, & Levine, 1994).
The lack of effective communication regarding EOL issues and the absence of EOL education were identified as significant factors contributing to poor care at the end of life (Edwards, Rogers, & Edmonds, 2003). In addition, perceived barriers to the completion of ACP among older adults were reported as perceiving ACP to be irrelevant, personal barriers, relationship concerns, information needs, health encounter time constraints, and problems with advance directive language (Schickedanz et al., 2009; Sudore et al., 2008). Most interventions have been based on educational perspectives emphasizing shared decision making (Fischer, Gozansky, Kutner, Chomiak, & Kramer, 2003; Holloran, Starkey, Burke, Steele, & Forse, 1995; Pearlman, Cole, Patrick, Starks, & Cain, 1995; Sulmasy et al., 1994). Some exceptions to this include work using health behavior change models to inform interventions of ACP (Fried, Bullock, Iannone, & O’Leary, 2009; Sudore et al., 2008). For example, Fried, Redding, et al. (2010) looked at how ACP behavior is a process over time and how individuals may be in different stages of the process. These stages include precontemplation (i.e., no intention to change behavior in the near future), contemplation (i.e., thinking about changing behavior in the near future), preparation (i.e., commitment to changing behavior soon), action (i.e., a recent change in behavior), and maintenance (i.e., ongoing behavior change).
One area in which the field shows some promise is targeted doctor-initiated communication with patients (Curtis, Patrick, Caldwell, & Collier, 2000; Hanson, Tulsky, & Danis, 1997). Some research shows that patients who reported having discussed their wishes for EOL care with doctors were more likely to receive care that was consistent with their preferences (Mack, Weeks, Wright, Block, & Prigerson, 2010). Doctors and patients in Curtis et al.’s (2000) study identified barriers to doctor-patient communication that could be overcome with education about EOL care, counseling interventions to overcome fears about dying, and health care systems changes such as providing more time and incentive for doctors to discuss EOL issues. The Study to Understand Prognoses and Preferences for Outcomes and Risks of Treatments found that 58% of severely ill, hospitalized patients did not want to have discussions about EOL care (SUPPORT Principal Investigators, 1995), but recent studies have found that the majority of patients wanted to discuss their EOL care decisions with their doctors but wait for the doctors to initiate the conversations (Dow et al., 2010; Perkins, 2007; Siegler & Levin, 2000). In consequence, there is growing consensus for multifaceted and coordinated interventions to promote doctor-initiated discussion of ACP in clinical encounters (Curtis et al., 2000; Field & Cassel, 1997; Teno, 2004; Teno et al., 2007).
The ACP interventions discussed above, although important, have neglected to give attention to two individual attributes that are worthy of study: religiosity and death attitudes. Older adults are often reluctant participants in EOL conversations because it involves communication about their fears related to their illness and desires for care and, perhaps most important, coping with common fears about death and dying (Curtis et al., 2000; Daaleman & Dobbs, 2010; Larson & Tobin, 2000). There are several individual-level factors that may contribute to death attitudes. For example, older adults with greater religiosity report less fear about death (Daaleman & Dobbs, 2010), but greater religiosity has also been significantly associated with wanting all measures to extend life in older adults with cancer (Balboni et al., 2010).
Although these attitudes in chronically ill older adults may decrease fear of death, little research has studied how religiosity and death attitudes may affect a person’s likelihood to engage in ACP. There may be a differential effect on the basis of factors that have been shown to be associated with engagement in ACP, such as demographics (e.g., age, educational background, race), physical and mental health function, social support, and physician-initiated communication about EOL care decisions (Fried, Redding, et al., 2010; Fried, Tinetti, et al., 2010; Sudore et al., 2010). In the current study, we examine the role these factors play in a sample of community-dwelling, chronically ill older adults.
Methods
Identification and Recruitment of Participants
Participants were drawn from a patient panel of chronically ill older adults: ambulatory care clinics from primary care practices participating in a practice-based research network in North Carolina. The North Carolina panel was identified from a cohort of patients who completed a health questionnaire in the waiting room prior to their doctors’ appointments. In this panel, participants aged 50 years or older who self-reported heart, liver, lung, kidney, or cerebrovascular disease were identified. Because chronically ill older adults constitute a small percentage of the community-based population (Lethbridge-Cejku & Vickerie, 2005), we lowered our inclusion age to increase our sampling panel. Potential participants were mailed a letter containing preliminary information about the study. The letter was followed up with a phone call that provided additional study details, verified eligibility criteria, obtained verbal consent, and scheduled a home interview for those who agreed to participate in the study. Exclusion criteria included lack of English facility, impaired orientation or memory loss, and non-community-dwelling status (e.g., nursing home residents).
Description of Study Participants
A total of 157 individuals participated in the study. The mean sample age was 69.4 years (SD = 10.9 years). Two thirds of the participants were women and White. Over half of the sample was married, and less than half had educations of high school or greater (see Table 1).
Characteristics and Outcome Statistics of the Study Sample (n = 157)
Note: ACP = advance care planning; DAP-R = Death Attitude Profile–Revised; GDS = Geriatric Depression Scale; OARS = Older Americans Resources and Services; PFI = Medical Outcomes Study SF-36 Physical Functioning Index.
Data Collection
Trained research assistants collected survey information in person from participants regarding health, functional, social, and psychological assessments; demographic characteristics; as well as an instrument that measured fear-of-death attitudes. A validated six-item screening tool (Callahan, Unverzagt, Hui, Perkins, & Hendrie, 2002) was used as a cognitive screening tool, and persons who had three or more errors were excluded. The study was approved by Institutional Review Board of the University of North Carolina at Chapel Hill.
Measures
ACP outcomes
Three separate dichotomous items (yes or no) were used as outcomes for three separate analyses. First was self-report of ACP discussions with a doctor. Participants were asked, “In the last six months, have you talked to your doctor about what kind of care you wanted if you were seriously or terminally ill or dying?” Second was self-report of ACP discussions with a family member. Participants were asked, “In the last six months, have you talked to your family about what your wishes were if you were seriously or terminally ill or dying?” Third was self-reported completion of a living will.
Social support
The amount of social support a person has can potentially affect his or her engagement in ACP. We measured social support using the Older Americans Resources and Services assessment of perceived dependability (Broadhead, Gehlbach, DeGruy, & Kaplan, 1988; Fillenbaum & Smyder, 1981). The measure has two dichotomous items (yes or no): (a) Do you have someone you can trust and confide in? and (b) Is there someone who would give you any help at all if you were sick or disabled?
Religiosity
Religiosity was measured using two items from the National Opinion Research Center, which were selected according to a previously tested and validated model of religiosity and health status (Davis & Smith, 1985; Levin, Chatters, & Taylor, 1995). These items—self-reported strength of religious or spiritual orientation and closeness to God (or a higher force)—were used to measure subjective or intrinsic religiosity. The scale for self-reported religiosity ranged from 1 (not at all) to 4 (strong). The scale for closeness to God (or a higher force) ranged from 1 (I don’t believe in God) to 4 (extremely close).
Health status
Self-rated functional and mental health status measures were included as factors that are known to be associated with engagement in ACP. The single-item measure of global health from the Years of Healthy Life Scale was used to determine self-reported health status. The Years of Healthy Life Scale consists of a self-assessment of general health (“In general, would you say your health is . . .”) and a 5-point, Likert-type response scale ranging from excellent to poor (Erickson, Wilson, & Shannon, 1995). The Physical Functioning Index of the Medical Outcomes Study SF-36 was used to measure functional health status (Stewart, Hays, & Ware, 1988). This 10-item instrument is a self-report of a range of severe and minor physical limitations (Ware & Sherbourne, 1992). The normative scale was used in which the raw score was divided by the maximum value of the scale and multiplied by 100 (Ware & Sherbourne, 1992). The index score ranges from 30 to 90. Mental health status was determined using the Geriatric Depression Scale (GDS) short form, a 15-item instrument with a dichotomous (yes or no) response format (Sheikh & Yesavage, 1986).
Demographics
The demographics included in the model were age in years, gender, race, and level of education. These were chosen on the basis of prior research that found these factors to influence ACP or completion of advance directives (Ko & Lee, 2010; McKinley, Garrett, Evans, & Danis, 1996; Meier et al., 1996; Morrison & Meier, 2004; Sudore et al., 2008, 2010). Racial categories in the study included White and African American. Education was dichotomized as college education (yes or no).
Death attitudes
We used two of the subscales from the Death Attitude Profile–Revised (Wong, Reker, & Gesser, 1994). The first subscale was Fear of Death. Fear of death pertains to a specific, conscious, and existential fear of the loss of self. The response format is a 7-point, Likert-type scale ranging from 1 (strongly disagree) to 7 (strongly agree), and a mean scale score is computed by dividing the total scale score by the number of items constituting each scale. The scale ranges from 1 to 7, where the higher the score, the more fearful of death. Some of the items in the measure include “I have an intense fear of death,” “The prospect of my own death arouses anxiety in me,” and “Death is no doubt a grim experience.” The Fear of Death subscale has a Cronbach’s α coefficient of .86 and 4-week test-retest reliability of .71. The second subscale used was Approach Acceptance. Approach acceptance connotes facing one’s own death with an associated positive affective reaction to this confrontation (Wong et al., 1994). The measure has 10 items using the same 7-point, Likert-type scale ranging from 1 (strongly disagree) to 7 (strongly agree), and a mean scale score is computed in the same manner as for the Fear of Death subscale. Some of the items in the measure include “Death brings a promise of a new and glorious life,” “I see death as a passage to an eternal and blessed place,” and “Death is a union with God and eternal bliss.” The Approach Acceptance subscale has a Cronbach’s α coefficient of .97 and 4-week test-retest reliability of .95.
Physician-initiated communication
We used a single dichotomous (yes or no) item to gauge whether a physician initiated communication with the respondent about EOL care decisions. The item asked, “In the last six months, has your doctor spoken to you about what kind of care you wanted if you were seriously or terminally ill or dying?”
Analyses
First, descriptive statistics were conducted on the three ACP outcomes and the variables in the model to check for missing data and outliers. Second, bivariate correlations were conducted to determine the most parsimonious regression models for the modest sample size. Third, three separate logistic regression models were conducted. All analyses were performed using IBM SPSS Version 19. Variables included in each of the three logistic regression models were those that had significant (p < .05) bivariate correlations with any of the three ACP outcomes. On the basis of these inclusion criteria, race, gender, and social support were excluded from the logistic regression models. Physician-initiated communication, although not significantly correlated in the bivariate analysis with two of the outcomes (i.e., discussions with family members and completion of a living will), was included in those two models, given its association with ACP (Dow et al., 2010). Physician-initiated communication was highly correlated with the ACP outcome of discussions with doctor (r = .534, p < .001), but it was excluded from this model because of the high likelihood that patients would report discussions with doctors if the doctors prompted the discussions, thus making the relationship tautological. A variance inflation factor test was conducted to test for multicollinearity. Variance inflation factor values that exceeded 2.5 for any of the independent variables, which is the recommended cut point for logistic regression models (Allison, 1999), were excluded from the models. On the basis of these criteria, approach acceptance and self-rated health were not included in the logistic regression models. The GDS measure had 15% missing data. The data appeared to be missing at random, so data were imputed for this variable using a regression method with the factors of age, sex, and education.
Results
The refusal rate for the sample was 26%. The descriptive statistics for characteristics and outcome measures for the sample are listed in Table 1. In addition to the demographics described earlier, the respondents had low levels of depressive symptoms, as reported by their GDS scores (M = 3.7, SD = 3.5) but had higher than average physical impairment scores according to the SF-36 Physical Functioning Index (M = 55.3, SD = 17.6). Close to half self-reported being in fair or poor health. They had high levels of perceived social support on the basis of the Older Americans Resources and Services dependence measure (M = 1.9, SD = 0.3). The mean score for the Fear of Death subscale was 2.82 (SD = 1.2), and the mean score for the Approach Acceptance subscale was 5.77 (SD = 1.0). The sample was strongly religious and close to God, with 90.3% reporting strong or somewhat strong religiosity and 65.6% reporting extreme closeness to God. For physician-initiated communication, 9.8% of subjects reported that their doctors had spoken to them in the past 6 months about what kind of care they wanted if they were seriously or terminally ill or dying.
Only 10.5% of respondents had talked to their doctors in the past 6 months about ACP, while more than half of the respondents (52.6%) indicated that they had talked with family members in the past 6 months about ACP. In reference to the completion of a living will, 38.3% reported that they had done so.
Logistic Regression Model Results
Results for the outcome of ACP discussions with doctors are highlighted in Table 2. Closeness to God was the only significant variable in the model and accounted for 29% of the variance; the model was significant (Nagelkerke’s R2 = .04, p < .01). Closeness to God had an odds ratio (OR) of 5.26; the closer respondents felt to God, they were 5.26 times more likely to have ACP discussions with doctors (95% confidence interval [CI] = 1.44 to 19.22, p < .05). The only other variable that approached significance was fear of death; for every unit increase in the fear of death score, there was 56% lower likelihood of having ACP discussions with doctor (OR = 0.44, 95% CI = 0.17 to 1.14, p = .09).
Stepwise Logistic Regression Model for Advance Care Planning Discussions With Doctor
Note: CI = confidence interval; GDS = Geriatric Depression Scale; PFI = Medical Outcomes Study SF-36 Physical Functioning Index.
p < .05. **p < .01.
The logistic regression model for ACP discussions with family yielded no significant factors, and the model was not significant. The only variable that approached significance was the mental health status variable of depression. For every unit increase in the GDS score, there was 10% lower likelihood of having ACP discussions with family (OR = 0.90, 95% CI = 0.80 to 1.02, p = .09; see Table 3).
Stepwise Logistic Regression Model for Advance Care Planning Discussions With Family
Note: CI = confidence interval; GDS = Geriatric Depression Scale; PFI = Medical Outcomes Study SF-36 Physical Functioning Index.
For the logistic regression model for the completion of a living will, fear of death explained 21% of the variance, and the model was significant (Nagelkerke’s R2 = .21, p < .01). The fear of death score had an OR of 0.66; for each unit increase in fear of death scores, there was 34% lower likelihood of having completed a living will (95% CI = 0.47 to 0.94, p < .05). The demographic variable of college education was very close to approaching significance. Those who were college educated were 2.13 times more likely to have completed a living will (OR = 2.13, 95% CI = 0.99 to 4.61, p = .054; see Table 4).
Stepwise Logistic Regression Model for the Completion of Living Will
Note: CI = confidence interval; GDS = Geriatric Depression Scale; PFI = Medical Outcomes Study SF-36 Physical Functioning Index.
p < .05. **p < .01.
Discussion
In a sample of chronically ill, older adults, we found that religiosity (i.e., closeness to God) was associated with ACP discussions with a physician. In addition, fear of death was associated with the completion of a living will. The finding that the closer one feels to God, the more likely one is to have engaged in discussions with a physician is important for clinicians to consider in future EOL decision-making interventions, particularly interventions that involve primary care physicians and geriatricians who have the most contact and communication with community-dwelling, chronically ill older adults and their surrogates (Fried, Tinetti, et al., 2010).
Prior research pointed to religion and other personal values as motivating factors for engaging in ACP (Cohen-Mansfield, Droge, & Billig, 1992) and with preferences for more life-prolonging treatment (Winter, Dennis, & Parker, 2009). Religion may ground values contributing to the cognitive process leading to the documentation of care wishes at the end of life. ACP is not a common focus within faith communities. In addition, such communities provide sets of worldviews to help guide value-based decisions, and many faith traditions have explicit guidelines and recommendations that address EOL situations and inform views on EOL care. Future research should explore faith communities as possible venues to improve ACP discussions.
Our finding of an association between fear of death and the completion of a living will is consistent with other work that has shown death fears such as denial of death to be an obstacle to planning for the EOL and the dying process (Cohen, McCue, Germain, & Woods, 1997; Zimmermann, 2007). Zimmerman (2007) stressed that among those who are in need of palliative care, denial of death can preserve hope, but “it becomes dysfunctional when the person needs to make decisions about real situations” (p. 302). Cohen et al.’s (1997) study of dialysis patients indicated that very few had completed advance directives and illustrated a “denial-like coping style” that obstructed preparation for death. Other work that has studied similar populations in North Carolina found that African Americans who were less comfortable discussing death were less likely to have completed advance directives (Johnson, Kuchibhatla, & Tulsky, 2008).
How death attitudes relate specifically to ACP discussions with a doctor has not been explored in previous research. The debate of how to tackle death fears on a personal and societal level has largely been ignored in the clinical medical literature (Zimmermann, 2007). Health practitioners working with chronically ill older adults should consider patients’ death fears and struggles with their own mortality as factors that may impede ACP discussions. Physicians should take into account their patients’ beliefs, such as faith traditions and personal philosophies, as resources to strengthen fear of death attitudes (Daaleman & Dobbs, 2010), which may enable patients to consider ACP discussions.
In the United Kingdom, educational intervention work to promote death awareness and ACP for those with chronic conditions has been conducted (Sanders, Rogers, Gately, & Kennedy, 2008). The qualitative analysis of participants in this program suggested that the death awareness and ACP section of a self-management course created a sense of disruption in the lives of those who generally try to use “positive thinking” or be hopeful in the face of living with a life-threatening illness (Sanders et al., 2008). The fact that ACP discussions may promote “negative thinking” or be disruptive is important to know when considering how to tailor future interventions that promote ACP discussions among chronically ill older adults.
Although physician-initiated communication was not significant in the two models in which it was included as a predictor, the descriptive information from this study is very telling. Only 10% of participants had talked with their doctors and fewer than 10% had their doctors talk with them about EOL preferences. Patients have expressed an interest in completing advance directives (Emanuel, Weinberg, Gonin, Hummel, & Emanuel, 1993), but many wait for physicians to initiate the discussion (Perkins, 2007). Curtis et al. (2000) identified barriers reported by doctors to initiating conversations about EOL care, which were related to lack of patient EOL education, the need for counseling about the fears of dying and needed health care system changes (e.g., lack of incentives and not enough time for the conversation to occur with current reimbursement system) (Curtis, et al., 2000). Doctors can be preoccupied with the management and control of chronic disease, and EOL issues may not be at the forefront of their minds. However, by the nature of having a chronic disease, this patient population is already at risk for future possible adverse events and is a target population to engage in the process of ACP by doctors. Furthermore, incentives for doctors to initiate EOL discussions with their patients need to be considered in health care policies, such as reimbursement for doctor consultations to determine goals for medical orders for EOL care.
There are a number of study limitations that need to be mentioned. First, the sample size was small, and the sample was drawn from primary care practices in North Carolina, and thus findings may not be generalizable to the larger population. A description of this primary care practitioner–based panel data by its developers (Sloane, Callahan, Kahwati, & Mitchell, 2006) reports that African Americans, Latinos, and older persons were enrolled at rates similar to the state’s 50-and-older population and that poor general health, chronic illness, and risk factors for chronic disease were more prevalent in the cohort than in the general 50-and-older community-based population. A larger, more representative sample of individuals would have reduced the chance of Type II error, which potentially hid the detection of smaller or more moderate statistically significant differences (e.g., fear of death in the first model, depression in the second model, and education in the third model).
Second, given that the study outcome of engagement in ACP was defined by whether individuals had completed a living will or had ACP discussions with family members or physicians in the past 6 months, our results may underestimate the number of individuals in the study who had engaged in other forms of ACP (e.g., completion of a durable health care power of attorney, do-not-resuscitate order, or do-not-hospitalize form), or had discussions with family members or their physicians beyond the 6-month time frame. The 6-month time frame was selected to avoid recall bias in discussions for family members. With regard to discussions with physicians, the 6-month time frame was based on clinical expertise by the research team that most community-dwelling older adults with chronic illness would have met with their physicians at least once within 6 months. Self-report about ACP could also have resulted in either an overestimation or an underestimation of actual ACP behavior. Finally, given that our data are cross-sectional, we cannot make causal assumptions.
In summary, among community-dwelling older adults with chronic illnesses, we found that religiosity and fear of death were associated with engagement in ACP when taking into account demographics, mental and physical health status, and physician-initiated communication. Given the significant findings, future interventions that involve increasing engagement in ACP should take into account a person’s religious background and explore older adults’ fears about death.
Footnotes
Acknowledgements
We are grateful to the patients who participated in this study. Also, we appreciate the careful review and comments made about this article by Brent Small, Sandy Reynolds, and Bill Haley from the University of South Florida, School of Aging Studies.
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
The author(s) received no financial support for the research, authorship, and/or publication of this article. This project was supported by the National Institute on Aging (Grant K23 AG01033).