Family Structure,Experiences with End-of-Life Decision Making,and Who Asked About Advance Directives Impacts Advance Directive Completion Rates

Patients and setting

We interviewed 130 inpatients in an urban university hospital. The hospital is set in a northeast environment and serves an urban and nearby suburban population with diverse racial, socioeconomic, and cultural backgrounds. Thirty-six percent of the patients admitted to this hospital are white, 59% black, and 5% of other backgrounds.

Approximately 47% of the patients are male and 53% are female. The hospital is nonspecialized, covering a wide variety of medical and surgical subspecialties. Based on a 6-month analysis of over 9800 billing codes for the hospital, approximately 65% of patients are admitted to a medical service, 16% to an obstetrics/gynecology service, 9% to a surgical service, 7% to the neurology service, and 3% to an oncology service.

Data collection

We collected several broad categories of data: patient demographics, health care utilization, disease entity, and experiences with death, dying, or end-of-life decision-making. Data were collected during one-on-one interviews and verified by chart review. Patients were chosen for interviews based on floor and bed number. The interviewers, who were internal medicine resident physicians, entered rooms at random on each hospital floor during interview days chosen based on the interviewers' individual schedules. Each interviewer entered approximately five rooms per floor per day. Interviewers did not have clinical relationships with the patients. Patients were excluded from the study if the bed was vacant at the time the interviewer entered the room, they were less than 18 years of age, required an intensive level of care, or were unable to speak English, communicate with the interviewer, or provide informed consent for any reason. Once informed consent was obtained, interviewers used a scripted survey to collect data regarding the presence of living wills or health care proxies, patient demographics, health care utilization, and experiences with death and dying. Health care utilization and disease characteristics included information about comorbidities, diagnoses, insurance status, chronicity of disease, presence of cancer, smoking history, visits to the primary care physician, and hospitalizations. A disease was categorized as acute versus chronic based on chart review by the physician interviewer. Acute disease was defined as a newly diagnosed illness and chronic disease as a progression or exacerbation of an ongoing underlying illness. Patients were asked if they had ever had previous discussions about advance directives and if so, who had initiated the conversation (e.g., physician, attorney, family, etc.). We gathered information about experiences with death and dying including having witnessed a cardiopulmonary arrest and having made end-of-life decisions for oneself or others. After the interview, medical charts were reviewed by the interviewer to confirm the information gathered and collect any additional data not provided by the patient.

Statistical analysis

We performed a bivariate analysis using Wilcoxon rank-sum and Fisher's exact tests to make comparisons between demographic and clinical characteristics of study subjects with and without living wills and health care proxies. We calculated counts and percentages or arithmetic means and standard deviations for all variables measured.

We conducted unadjusted logistic regression analysis to evaluate the association of potential covariates with possession of advance directives. We next conducted a multivariable analysis considering potential covariates. In the initial multivariable model, we included covariates that had a p value <0.10 in the univariate analysis. We used backward elimination techniques for variable reduction, and retained covariates when they were significant or when removal caused a greater than 10% change in the estimate for possession of a living will or health care proxy. A biostatistician performed the statistical analysis. We obtained approval from the Institutional Review Board prior to subject recruitment.

Demographics and advance directives

We interviewed a total of 130 inpatients. Patient demographics, disease characteristics, and prevalence of living wills and health care proxies are shown in Table 1. Survey participants were mostly female (57%) and black (65%). Sixty-four percent of patients were admitted for nononcologic, medical diagnoses (Table 1). Our study had an equal mix of acute and chronic disease processes as the reason patients' admission to the hospital. Eighty-two percent of our population had attained educational status of high school or less and 95% percent had health insurance. Ninety percent of patients reported a religious affiliation.

Bivariate analysis

Table 2 shows the results of the bivariate analysis comparing those with and without a living will. There was no difference in race or gender in patients who possessed a living will. Older patients were more likely to possess a living will (p=0.0046). Patients with living wills had more comorbidities than those without living wills (p=0.018). Measures of health care utilization such as weeks since last primary care physician (PCP) visit and number of admissions in the last year did not impact possession of living wills. Of our hospitalized patients with a living will, 74% were admitted for a chronic condition, whereas those without living wills were equally likely to have an acute or chronic condition as the reason for their admission (p=0.009). Admission with an acute disease process did not show similar proportions. Most patients in our study were married, however, the proportion of those with a living will who were widowed was significantly higher than those who did not have a living will (p=0.02). Experience with death and dying of a loved one or witnessing a cardiopulmonary arrest did not affect possession a living will.

The results of the bivariate analysis for patients with and without a health care proxy are displayed in Table 3. Older patients were more likely to possess a health care proxy (p≤0.001) and there was no difference in race or gender. Those with a health care proxy were more likely to report a religious affiliation than those without a health care proxy (p=0.04). Measures of health care utilization such as weeks since last PCP visit and number of admissions in the last year did not impact health care proxy possession.

Patients with a health care proxy had more children than those without one (p=0.03). Those with health care proxies were four times more likely to be widowed than those without a health care proxy. The majority of patients without health care proxies were single (p<0.001).

Among those who had a health care proxy, 35.6% had made end-of-life decisions for another while only 15.3% had done so in the group without a health care proxy (p=0.01). Experience with death and dying of a loved one or witnessing a cardiopulmonary arrest did not affect possession a health care proxy.

Logistic regression analysis

The results of the logistic regression analysis adjusting for age, race, and marital status are displayed in Table 4. Patients who were asked about advance directives by medical staff were 10.8 times more likely to complete a living will (95% confidence interval [CI] 4.59–25.3) and 1.68 times more likely to complete a health care proxy (95% CI 0.81–3.47) than those who were not asked. Those asked by legal staff were 46.5 times more likely to have a living will (95% CI 15.5–139.4) and 4.34 times more likely to have a health care proxy (95% CI 1.5–12.6). Those asked by family or friends were 68.6 times more likely to have a living will (95% CI 13.0–361.3) and 18 times more likely to have a health care proxy (95% CI 2.03–158.8).

Patients who had experience with end-of-life decision-making were 2.54 times more likely to possess a living will than those who had not made end-of-life decisions (Table 4, 95% CI 1.01–6.42). Similarly, patients with decision-making experiences were 3.53 times more likely to possess a health care proxy (95% CI 1.51–8.25).

Patient demographics

It is well established that patient age is associated with higher rates of advance directive completion.^5,9,11 Our study confirmed that older patients are more likely to complete either a living will or a health care proxy. Although not a predictor for possession of a living will or health care proxy, we found that 97% of the patients with a health care proxy had a religious affiliation, compared to 86% of patients without a health care proxy. This difference was statistically significant. Previous research has suggested relationships between religiousness and end-of-life preferences. ¹² Further study is needed to delineate the implications of this. We hypothesize that religiously affiliated individuals may have increased exposure to forums that enable end-of-life conversations, such as church or synagogues.

Family structure

We found that number of children and marital status impacts possession of advance directives. Specifically, patients with health care proxies had more children than those without. This result was not true of living wills nor did it hold true after adjustment for age, marital status, and gender. A study by Greenberg et al. ⁴ found that having children was a strong predictor of having a living will. That study looked solely at the presence or absence of children, whereas we quantified the number of offspring to see whether that had an impact on advance directive possession. Our findings could suggest that those with multiple children are encouraged to designate a single person as the surrogate decision-maker, which is done when completing a health care proxy but not necessarily a living will.

While most of our patients were married, those with living wills or health care proxies were more often widowed than those without them. This furthers our hypothesis that experiences with death and dying, such as one's spouse, might motivate patients to consider and engage in completing an advance directive. Taken together, our findings suggest that family structure has influence on advance directive completion and warrants further study.

Health care utilization

We hypothesized that a patient's health care utilization would impact rates of advance directive completion. We generated this hypothesis based on the PSDA mandate requiring hospitals to provide information about advance directives. Using the time since last visiting a PCP and number of hospitalizations as measures of health care utilization, we found that increased interaction with health care did not impact possession of a living will or health care proxy. Simply being asked about presence of advance directives by a medical provider increased the likelihood that of possession by 10.8 times for living wills and 1.68 times for health care proxies. This is an interesting finding considering that the efficacy of the PSDA in increasing utilization of advance directives has been questioned. ¹³ A 2007 study found that 27% of admitted cardiac patients did not recall having a discussion about advance directives as mandated by the PSDA. ⁷ It is possible that patients did not recall being asked the question because they were distracted by the admission process or by their acute illness. In addition, hospitals often have clerks ask about advance directives during the admission process; this practice, although it is in compliance with the PSDA, is unlikely to impact completion of advance directives.

Our finding that hospital admissions did not improve advance directive possession offers further evidence of the ineffectiveness of the PSDA, although it is possible that, despite the PSDA, patients were not asked to complete an advance directive while admitted. While we found medical providers asking patients about advance directive increased the likelihood of completion of both living wills and health care proxies, our interviews did not categorize the setting in which this conversation occurred.

Overall, our study suggests that our chosen measures of health care utilization are not a good marker of possession of advance directives, but that the simple process of asking a patient about an advance directive has a profound impact on completion rates. These data should reassure clinicians that the topic of advance directives is an important one to bring up during clinical encounters. Larger, more generalizable studies are needed to confirm all of these findings.

Disease process

As we expected, our study showed that patients with living wills had more comorbidities than those without them. A study by Thorevska et al. ⁶ showed that patients with higher APACHE (Acute Physiology and Chronic Health Evaluation) scores, a measure of disease severity, were more likely to possess a living will. Our study did not include measures of disease severity.

We found that disease chronicity impacts advance directive possession. Patients admitted for reasons related to a chronic disease were more likely to possess a living will than those who were admitted as a result of an acute process. Lamont and Siegler ¹⁴ found that being admitted emergently to the hospital in itself was not associated with an increased willingness to discuss advance directives, but that 50% of patients with emergency admission specifically due to dyspnea desired to discuss advance directives. In comparison, only 18% of patients with other admitting symptoms wished to discuss them. ¹⁴ When considered with our findings, this suggests that acute exacerbations of chronic diseases encourage people to consider advance planning and can serve as a trigger for completing an advance directive.

It is reasonable to assume that not only disease chronicity but also disease entity would impact the completion of advance directives, and previous studies have supported this notion.^3,7,14 Further evidence that advance care planning is disease-specific comes from a study comparing patients with cancer to cardiac patients. ⁷ The study found that patients with advanced cardiac disease were more likely to indicate interest in receiving information about advance directives than those with cancer. ⁷ Others have suggested that goals of treatment may impact likelihood of completing advance directives. ² Although we did not find that the diagnosis of cancer itself impacted advance directive completion, we demonstrated that chronic disease increases the likelihood of having a living will. Thus, disease characteristics and disease entity do seem to influence willingness to discuss and perhaps complete an advance directive.

End-of-life experiences

We found that patients who had made end-of-life decisions for another were more likely to go through the process of preparing an advance directive, even when adjusted for age, gender and marital status. Patients who had made decisions for another were 2.5 times more likely to possess a living will and 3.5 times more likely to possess a health care proxy. Simply having a close friend or relative die, however, did not predict possession of an advance directive in our study. Greenberg et al. ⁴ found that one of the strongest predictors for having an advance directive was having faced a life-threatening experience oneself.

Our findings may indicate that the decision-making process, rather than being close to a dying experience or witnessing a cardiopulmonary arrest affects patients in such a way that they go on to complete an advance directive for themselves. However, we did not deeply probe the patients' end-of-life experiences and cannot be sure whether these experiences had an impact on completion, or if they occurred before the advance directive was completed. We do show that being a decision maker for someone else impacts completion of advance directives and should be the topic of further investigation.

Finally, we found that the person asking a patient about advance directives has an impact on completion. When medical staff asked the patient, they were 10.8 times more likely to complete an advance directive; this number rose to 46.5 for legal staff and even higher at 68.6 times more likely for family and friends. Similar trends were seen in regards to health care proxies. Alano et al. ⁹ found that 81% of patients who had an advance directive had been asked about advance directives and received an explanation of why they were important. Similar to our findings, researchers in that study found that the most common person to have asked the patient about advance directives was their child. ⁹

In our population, lawyers were more successful than medical staff in getting patients to complete a living will or health care proxy. The study by Alano et al. ⁹ reported that in an older population, health care providers were the primary influence in completing advance directives only 8% of the time. In that study, lawyers had less influence on completing advance directive (6%) and children had the greatest influence (26%). ⁹ In our population, physicians most commonly asked patients about advance directives but were far less successful than lawyers. Studies investigating these findings are warranted.

Living will versus health care proxy

Our findings suggest that while there are important similarities and differences in regards to living wills and health care proxy completion. While both of these documents fall under the umbrella of advance directives, we found in our bivariate analysis that different factors influence whether a person completes a living will or health care proxy. For example, we have shown that living will but not health care proxy completion is increased based on comorbidities and disease entity. Health care proxies, however, were not impacted by these factors. Conversely, religion, and number of children affected health care proxy completion but not living will. This variation has been seen in other studies as well.^4,5 For example, racial differences in advance care planning were less pronounced with health care proxies than with living wills. ⁵

Limitations

Our results should be interpreted with caution. First, our sample size was small and our population was predominantly female and black. Although historically demographics such as age, race, and gender have been shown to have an impact on advance directives and end-of-life preferences, our study did not confirm these findings, which may reflect differences in our patient population. Second, we had a low rate of advance directive completion, although our rates are consistent with those reported in the national literature. Finally, several of our variables were based on patients' recollections about their experiences and may be inaccurate, allowing for a reporting bias.