Disease Specific Advance Care Planning for Heart Failure Patients: Implementation in a Large Health System

Abstract

Purpose:

This study evaluates the implementation of Disease Specific Advance Care Planning (DS-ACP) for heart failure patients in a large health care system. The DS-ACP model uses a trained facilitator to guide communication between patient and proxy regarding patient values and goals for treatment in worst-case scenarios, develop a disease specific documentation plan, and identify needed resources.

Methods:

Referral and DS-ACP service delivery data were merged with electronic health record data. Process measures for implementation are described. Patients who participated in DS-ACP are compared with those who did not on demographics, medical characteristics, and outcomes such as inpatient readmissions, and hospice use among those who died. A chart audit was completed on a sample of patients to examine differences in advance care planning documentation between participants and nonparticipants.

Results:

Nearly one third (31.8%) of the 1894 heart failure patients referred to DS-ACP participated. Referral method was associated with participation. A chart audit revealed 94.3% of those completing the DS-ACP process, had a health directive compared to 24.8% of noncompleters (p<0.001). Of the patients who died by the end of the study period (n=286), DS-ACP participants were more likely to have used hospice compared to nonparticipants (56% versus 37%, p=0.002).

Conclusions:

These initial evaluation results indicate the DS-ACP model, previously tested in randomized trials, can be successfully implemented among heart failure patients in a large health care system. Results demonstrate the importance of referral method and show initial positive results for participation, improved documentation, and hospice use.

Introduction

The Patient Self-Determination Act is now 20 years old, yet there is much to be learned about the type of assistance patients require to create their own written advance directives. Advance care planning (ACP) models have emerged to promote a process of communication in helping patients make informed choices, and the creation of reliable systems to honor these choices. This emerging model of ACP, however, must be individualized to the patient's readiness to learn and stage of illness. The majority of deaths in individuals over age 65 are the result of chronic disease, yet less than 50% of these patients are likely to have created written plans.¹ This is unfortunate as these patients often experience unexpected periods of health crisis and recovery with no clear trajectory guiding when end-of-life decisions will be needed.¹

Conventional advance directives (e.g., living will, power of attorney for health care) are not constructed to address the specific planning needs of patients with chronic disease. Few people have such directives, physicians are often unaware of their presence due to poor accessibility, and the language in directives is often too generic focused around life sustaining treatment to be helpful,¹ particularly for patients with chronic or terminal illness.^2,3 Additional barriers to effective ACP include poorly prepared surrogates and communication gaps between patients and providers on the end-of-life decision-making. The appointment of a health care agent is a recommended strategy in ACP, but research has found that named surrogates were often not present to make decisions, were too emotionally upset to offer guidance,¹ and most importantly, were frequently unsuccessful at accurately predicting the desired choices of the patient.^4,5 Evidence also suggests many medical professionals are too uncomfortable or lack appropriate training to initiate complex end-of-life decision-making discussions.⁶

Disease Specific Advance Care Planning (DS-ACP)⁷ represents a novel, tailored approach to facilitate ACP in chronically and terminally ill patients. The DS-ACP model, which incorporates a facilitated planning process and emphasizes condition-specific treatment scenarios, has been examined in small randomized trials for conditions such as congestive heart failure, end stage renal disease,⁷ human immunodeficiency virus/acquired immune deficiency syndrome (HIV/AIDS),⁸ and heart surgery.^7,9 These early trials demonstrate that patients and proxies are receptive to the intervention, and that the DS-ACP model results in an improved understanding of patient goals among intervention group surrogates. Larger randomized trials have confirmed that among DS-ACP participant pairs, surrogates acquire a significantly higher degree of understanding of patient goals,¹⁰ end-of-life wishes were fulfilled more frequently, and family members experienced less stress anxiety and depression.¹¹ This process fits with clinical guidelines calling for providers to complete a systematic review and document desired goals for care among patients with potentially life-limiting or chronic progressive illnesses.¹² The purpose of this article is to describe initial outcomes for the first 2 years implementing DS-ACP for heart failure (HF) patients in a large health care system. In contrast to prior randomized trials, this study provides an evaluation of DS-ACP in an applied setting.

Methods

Subjects

Study subjects were patients with a primary or secondary HF diagnosis served by Allina Hospitals & Clinics, a large health system consisting of 11 hospitals and over 70 primary care clinics located in Minnesota and western Wisconsin. From 2008–2009, 2029 HF patients were referred for DS-ACP. The current study is limited to 1894 of those patients referred in the first 23 months to allow enough time for postreferral events (e.g., participation in DSACP, rehospitalization) before linkage with medical records to collect outcomes (mid-February 2010). Additionally, 79 cases were excluded due to incorrect medical record number in the referral database. Patients selected for this study were confirmed to have a HF diagnosis in the electronic health record (EHR). The Allina Institutional Review Board approved the study.

Intervention

The Respecting Choices DS-ACP program developed by Gundersen Lutheran, La Crosse, Wisconsin, was selected for implementation with the HF population.¹³ The DS-ACP program was designed to address the specific planning needs of patients who have advanced chronic illness, such as those experiencing complications from their disease, or more frequent hospitalizations. This program includes (1) a facilitated in-depth and disease specific ACP interview with patients and their proxy, (2) certified facilitators who successfully complete a 26-hour competency-based communication skills training program, and (3) disease-specific planning tools to accurately document patient goals, values, and treatment preferences that are accessible in the medical record. A brief description of the DS-ACP interview is included in Appendix A.

Three referral strategies were used to reach HF patients across the system from primary care, inpatient, and home care and case management. First, the discharge order set for HF was modified to include a DS-ACP referral. Second, a list of HF patients was provided to primary care physicians, asking them to identify patients they thought might benefit from DS-ACP. Letters were sent to patients from the physicians to introduce DS-ACP. This referral strategy was only completed with two clinics due to time constraints. The third referral strategy was to generate direct referrals from medical providers and care managers. Education on DS-ACP and how to make referrals was targeted to those serving HF patients through hospitals and home care programs.

The DS-ACP program was implemented by Allina's Home and Community Service's Palliative Care Department with eight staff (6 registered nurses and 2 social workers) serving as facilitators, and a central referral tracking and dispatch role. Referrals were received at the care navigation help desk through electronic and telephone orders or patient inquiry. The referral coordinator called patients to follow up on referrals, explain the ACP process, set up appointments, and assign patients to a facilitator. Coordinators made several attempts to reach patients and left up to three messages.

Facilitators reviewed the patient's medical record before the session to understand the patient's stage of health. Post session, the facilitator made copies of completed documents scanned them into the EHR, and returned the originals to the patient and proxy. A summary of the session was documented in the EHR in addition to a follow up with the provider. This summary included patient wishes as well as information about identified needs to be addressed by referrals for other services or through further discussion between patient and provider.

Data collection and measures

The Palliative Care Department maintained a database with referrals to DS-ACP in order to track referral source, calls to patients, document interest in or refusal of DS-ACP, assign facilitators, track appointments, appointment failures and completion. Records from the referral database were linked with EHR data on demographics, key diagnoses, hospitalization, death, and hospice use. Records were pulled through February of 2010. As referrals to the program occurred throughout the 2-year period, not all patients contributed equal time to the study period. Patients were considered noncompleters if they had not done DS-ACP by the end of the follow-up period. Chart audits were conducted on a randomly selected sample of 105 DS-ACP participants and 105 nonparticipants for the purpose of examining ACP documentation in charts. Audits consisted of examining the EHR for any scanned documents related to ACP and indicating if the patient had any of the three types of documents in their chart.

Process measures include number of referrals, referral source (discharge order; clinic or hospital provider; care coordinators, home care clinicians, or other case managers; letters sent from the patient's clinic), and patient uptake of DS-ACP intervention (yes/no). Outcome measures include types of documentation in the EHR (resuscitation guidelines, advance directive, statement of treatment preferences (STP), or physicians orders for life-sustaining treatment (POLST), and inpatient hospitalization (yes/no) or emergency department admission (yes/no) within 30 or 60 days after referral. For those who died during the study period additional outcome measures include hospice use (yes/no) and hospice length of stay (days). Additional measures used to describe patients and compare those who did and did not participate include age (grouped as younger than 65 years and 65 years or older), race (African American, American Indian, Asian, Hispanic, white), gender, martial status (married, single, widowed, divorced/separated), and selected other disease diagnoses (chronic kidney disease including renal failure, chronic obstructive pulmonary disease [COPD], diabetes).

Analysis

Participants and nonparticipants were compared on patient and referral characteristics using frequencies and proportions, with χ² tests. Prevalence of EHR documentation was also compared among random samples of participants (n=105) and non-participants (n=105) using proportions and χ² tests.

A subanalysis was conducted among patients who had been referred to DS-ACP as a result of a hospital discharge (n=859) to examine rehospitalization at 30 and 60 days after the discharge order. All patients in this group had been referred at least 60 days prior to the data extract to ensure all subjects remain in the denominator when examining 60-day readmission rate. These patients were classified into three groups: (1) completed DS-ACP within 30 days of referral; (2) completed DS-ACP 30 or more days after discharge; and (3) did not complete DS-ACP. Cross-tabulations and χ² tests were used to compare the frequency of readmissions across the three categories of patients.

Further subanalysis examined patients who had died between the beginning of the referral period and the collection of outcome measures. Among patients who had died (n=286), χ² tests were computed to compare the difference in hospice use between DS-ACP participants and nonparticipants. Logistic regression models were run to examine hospice use, adjusted for differences in age and referral method. The sample was further restricted to hospice users (n=125) to examine differences in mean and median hospice length of stay (LOS). Hospice LOS was not normally distributed, so the Mann-Whitney two-sample test was used as a nonparametric alternative to assess difference in means between participants and nonparticipants. An additional test for the equality of medians was also computed.

Results

Of the 1894 HF patients referred, 602 (31.8%) completed DS-ACP. Table 1 presents selected characteristics of all HF patients referred for DS-ACP by participation status. The majority of patients were referred during the second year. Those referred during the second year were less likely to have completed DS-ACP at the end of the study period, likely due to less follow-up time. Nearly all were referred only once, and the most common referral method was the discharge order. Type of referral was significantly associated with participation in DS-ACP (p<0.001). Referrals by medical provider (40% versus 25.4%) or case managers and home care workers (30.1% versus 14.4%) were associated with increased DS-ACP participation. Conversely, those referred through discharge orders represented a smaller proportion of participants compared with non-participants (23.9% versus 54.7%, respectively). There was no association detected between DS-ACP completion and gender, race, marital status. Those completing DS-ACP were an average of 1.5 years older than those not completing (75.63 years versus 73.84, p=0.017), and a small association was found with diabetes. A majority of sessions (77%) were conducted at patient's homes, the rest were conducted in clinics (15%), nursing homes (3%) or hospitals (2%).

Table 1.

Selected Characteristics of Patients Referred for DS-ACP by Completion Status, January 2008–November 2009

	Patients completing DS-ACP (n=602)		Patients not completing DS-ACP (n=1292)		All patients referred for DS-ACP (n=1894)
Referral Information	Frequency	%	Frequency	%	Frequency	%	p value^a
Year of referral
2008	244	40.5	309	23.9	553	29.2	<0.001
2009	358	59.5	983	76.1	1341	70.8
Number of times patient was referred
1	556	92.4	1184	91.6	1740	91.9	0.852
2	42	7.0	98	7.6	140	7.4
3–4	4	0.7	10	0.8	14	0.7
Referral method
Discharge order	144	23.9	707	54.7	851	44.9	<0.001
Clinic or hospital provider	241	40.0	328	25.4	569	30.0
Care Coordinators/Home Care	181	30.1	186	14.4	367	19.4
Clinic letter	33	5.5	65	5.0	98	5.0
Clinic letter and discharge order	3	0.05	6	0.5	9	0.5
Gender
Male	288	47.8	667	51.6	955	50.4	0.111
Female	313	52	625	48.4	938	49.5
Unknown	1	0.2	0	0.0	1	0.1
Race
White	537	89.2	1165	90.2	1706	87.4	0.377
African American	27	4.5	52	4.0	79	4.0
Asian	4	0.7	19	1.5	23	1.2
Hispanic	7	1.2	13	1.0	20	1.0
American Indian	5	0.8	13	1.0	18	0.9
Unknown	22	3.7	30	2.3	105	5.4
Marital status
Married	287	47.7	643	49.8	930	49.1	0.159
Widowed	196	32.6	380	29.4	576	30.4
Single	79	13.1	163	12.6	242	12.8
Divorced/separated	35	5.8	102	7.9	137	7.2
Unknown	5	0.8	4	0.3	9	0.5
Age
<65	102	16.9	291	22.5	393	20.7	0.003
65+	492	81.7	995	77.0	1487	78.5
Other chronic conditions^a
Chronic kidney disease	235	39.0	555	43.0	790	41.7	0.110
COPD	146	24.3	289	22.4	435	23.0	0.379
Diabetes	236	39.2	570	44.1	806	42.6	0.046

For χ² test.

Chronic conditions are not mutually exclusive.

Table 2 shows the results of a chart audit to examine the presence of different types of ACP documentation in the EHR at the end of the study period. Resuscitation guidelines were not significantly different between the two groups. Of those completing the DS-ACP process, 94.3% had a health directive compared to 24.8% of noncompleters (p<0.001). A patient could have a health directive completed either through DS-ACP or outside of that process. Health directives were classified as designating a proxy only or designating a proxy as well as indicating wishes (often these were resuscitation guidelines). While both types of health directives were more prevalent among patients who had completed DS-ACP, the biggest difference is seen in directives that just designate a proxy. This is likely because the DS-ACP process relies on the Statement of Treatment Preferences (STP) to document specific desires. The STP form is specific to the DS-ACP process but the audit looked for any documentation that was specific to the symptoms and treatment of HF. Only patients going through the DS-ACP process had documentation specific to their medical condition (84.8% of completers had a STP compared to 0 noncompleters, p<0.001). POLST forms were present for 3.8% of DS-ACP patients. This is likely lower than would be expected because the POLST form was not introduced until the second year of the project and then was used initially only among hospice patients. The chart audit found that 3% of patients who had completed the DS-ACP interview had no documentation. These patients did in fact participate in the full interview with their proxy to talk about their preferences but declined the option to create documents at the end of the session.

Table 2.

Chart Audit Results to Assess Documentation in the Electronic Health Record

Type of Documentation	Completed DS-ACP (n=105) %	No DS-ACP (n=105) %	p value
Resuscitation guidelines	17.1	12.4	0.331
Health directive
Designating proxy only	61.9	2.9	<0.001
Designating proxy and indicating wishes	32.4	21.9	0.088
Any health directive	94.3	24.8	<0.001
Physician Orders for Life Sustaining Treatment (POLST)	3.8	0	0.043
Statement of treatment preferences/or disease-specific documentation	84.8	0	<0.001

DS-ACP, Disease Specific Advance Care Planning.

Table 3 shows inpatient and emergency department admissions 30 and 60 days after discharge for patients who were referred to DS-ACP through a HF discharge order. There were no significant differences in readmissions.

Table 3.

Inpatient and Emergency Department Admissions Among Patients Referred for DS-ACP During a Prior Hospitalization (n=859)

Admissions	Completed DS-ACP within 29 days of Discharge order (n=81) %	Completed DS-ACP 30+ days after discharge order (n=63) %	No DS-ACP (n=715) %	Total patients with discharge order (n=859) %	p value
Inpatient admission within 30 days of referring discharge	17.3	23.8	22.7	22.2	0.519
Inpatient admission within 60 days of referring discharge	28.4	34.9	31.5	31.4	0.704
ED admission within 30 days of referring discharge	9.9	7.9	7.4	7.7	0.730
ED admission within 60 days of referring discharge	12.3	17.5	13.0	13.3	0.587

DS-ACP, Disease Specific Advance Care Planning; ED, emergency department.

Among the 286 deceased patients, 56.1% of those who had completed DS-ACP were enrolled in hospice compared to 37.2% of those who had not completed DS-ACP, p=0.002 (Table 4). Logistic regression was conducted to adjust for the differences seen in age and recruitment method between DS-ACP completers and noncompleters in univariate analysis. After adjustment for these differences, the patients who had completed DS-ACP were still over twice as likely to have participated in hospice than noncompleters (odds ratio [OR]=2.21, p=0.003, confidence interval [CI]=1.3, 3.7). Among those who used hospice, mean LOS differed by 27.5 days between DS-ACP participants and non-participants (71.4 and 43.8 days, respectively). Likewise, median LOS differed by 11 days between participants and non-participants (24 and 13 days, respectively). While the difference in these between groups appears clinically relevant, our sample size was underpowered to achieve statistical significance.

Table 4.

Comparison of Hospice Use among Patients Who Died During Study Period (n=286)

	Completed DS-ACP (n=98) %	No DS-ACP (n=188) %	Total (n=286) %	p value
Used hospice	56.1	37.2	43.7	0.002
Hospice patients only	(n=55)	(n=70)	(n=125)
Mean (SD) hospice length of stay (days)	71.4 (107.6)	43.8 (62.2)	56.0 (85.9)	0.110
Median hospice length of stay (days)	24	13	19	0.187
Interquartile range (25th–75th)	5–88	5–62	5–75

DS-ACP, Disease Specific Advance Care Planning; SD, standard deviation.

Discussion

This study demonstrates that the Respecting Choices DS-ACP model previously tested in randomized trials can be implemented into a large health system as a service for patients with chronic, progressive illness. Resources critical to the success of the first 2 years of DS-ACP implementation included trained facilitators with dedicated time, a call center to track referrals, staff time to train others in the system about DS-ACP and the referral process, and EHR system for documentation.

The study provides insight about the best referral strategies with those referred through a medical provider or care coordinator being most likely to participate. This is likely related to their ability to explain the DS-ACP process and what to expect in the interview. The study also identified improved documentation of wishes for DS-ACP participants and provides evidence that DS-ACP participation is associated with increased hospice use. While documentation is essential for being able to honor patient's wishes, another key element is the education of proxies to make those decisions. Prior studies on the Respecting Choices DS-ACP model have demonstrated, through randomized trials, this process of ACP is effective at increasing agreement between proxy decisions and patient wishes.^7–10 This study adds to these findings the experience of implementing DS-ACP in a large health care system and some positive initial outcomes related to patient wishes documentation, and patient care experience. A randomized trail in Australia with elderly inpatients found positive results for the model at the point of end-of-life care decision making and for family experience.¹¹ The experience of care decisions made by providers and proxies after completing the DS-ACP process is an area that warrants further research for those with chronic diseases.

While the evaluation of this early implementation phase shows promising results the study is subject to some limitations. As with other studies using health record codes to identify HF, our study is subject to biases that may result from inaccurate coding of HF. The small sample size resulted in lack of power to examine hospital readmissions and hospice LOS. The study also only had information on readmissions within our system hospitals. It is possible that relationships identified between DS-ACP and outcomes of hospice use or readmissions were influenced by other factors that cannot be adjusted for here. However, the similarity of participants and non-participants and the large difference in hospice use and documentation between the two groups are indications of positive outcomes of the intervention.

Leaders of the DS-ACP implementation effort encountered some challenges that could be experienced by others trying to replicate the model. The first was creating culture change needed to get some providers to support DS-ACP and to see how ACP is different than the current state of completing health care directives. Some felt the directives were sufficient, some felt they were not qualified to have conversations with patients even for referrals, others felt only doctors should have the conversations with patients rather than the facilitators. To address the culture change needed, the DS-ACP team engaged system senior leadership who adopted DS-ACP as a quality improvement initiative. The team also consistently made the case for change by relating it to quality of care and the patient/family experience. A second challenge was the process of engaging patients in the ACP process. Education materials were developed including a video and ACP discussion guide for providers. Staff and providers needed different levels of training to be able to talk with and engage patients in the process. A large-scale training with different levels of information and skill building was created and implemented after this study period. A third barrier was the challenge of staff/provider inability to view ACP documentation and summary of ACP discussions in the EHR as notes and documents can be somewhat buried in the system. A centralized EHR component was developed so ACP information is easily visible in all care settings.

In conclusion, the resources devoted to implement DS-ACP proved to be a worthy investment providing positive initial findings about participation, referral strategies, hospice use, and documentation. These findings and initial lessons led to the development of a comprehensive, tiered training and an EHR component. These tools are being used as the system spreads basic ACP and DS-ACP across other clinical communities in the health system.

Footnotes

Author Disclosure Statement

No competing financial interests exist.

Appendix A

References

Kass-Bartelmes

, Hughes

, Rutherford

. Advance Care Planning: Preferences for Care at the End of Life. Research in Action. Rockville, MD: Agency for Healthcare Research and Quality, 2003.

Messinger-Rapport

, Baum

, Smith

. Advance care planning: Beyond the living will. Cleve Clin J Med, 2009; 76:276–285.

Lynn

, Goldstein

. Advance care planning for fatal chronic illness: Avoiding commonplace errors and unwarranted suffering. Ann Intern Med, 2003; 138:812–818.

Hines

, Glover

, Babrow

, Holley

, Badzek

, Moss

. Improving advance care planning by accommodating family preferences. J Palliat Med, 2001; 4:481–489.

Uhlmann

, Pearlman

, Cain

. Physicians' and spouses' predictions of elderly patients' resuscitation preferences. J Gerontol, 1988; 43:M115–121.

Weissman

, Ambuel

, Norton

, Wang-Cheng

, Schiedermayer

. A survey of competencies and concerns in end-of-life care for physician trainees. J Pain Symptom Manage, 1998; 15:82–90.

Briggs

, Kirchhoff

, Hammes

, Song

, Colvin

. Patient-centered advance care planning in special patient populations: A pilot study. J Prof Nurs, 2004; 20:47–58.

Lyon

, Garvie

, McCarter

, Briggs

, He

, D'Angelo

. Who will speak for me? Improving end-of-life decision-making for adolescents with HIV and their families. Pediatrics, 2009; 123:e199–206.

Song

, Kirchhoff

, Douglas

, Ward

, Hammes

. A randomized, controlled trial to improve advance care planning among patients undergoing cardiac surgery. Med Care, 2005; 43:1049–1053.

10.

Kirchhoff

, Hammes

, Kehl

, Briggs

, Brown

. Effect of a disease-specific planning intervention on surrogate understanding of patient goals for future medical treatment. J Am Geriatr Soc, 2010; 58:1233–1240.

11.

Detering

, Hancock

, Reade

, Silvester

. The impact of advance care planning on end of life care in elderly patients: randomised controlled trial. BMJ, 2010; 340:c1345.

12.

Institute for Clincial Systems Improvement. Health Care Guideline: Palliative Care, Third. Bloomington, MN: Institute for Clincial Systems Improvement, November 2009.

13.

Kirchhoff

, Hammes

, Kehl

, Briggs

, Brown

. Effect of a disease-specific planning intervention on surrogate understanding of patient goals for future medical treatment. J Am Geriatr Soc, 2010; 58:1233–1240.