Life‐Sustaining Treatment Documentation in VA Home Based Primary Care Improves With Feedback Reports and Facilitation

ABSTRACT

Background

The Veterans Health Administration (VHA) Life‐Sustaining Treatment Decisions Initiative (LSTDI) aims to improve documentation of patient preferences for life‐sustaining treatment (LST), particularly in high‐risk populations such as those served by Home Based Primary Care (HBPC) programs. Despite this mandate, LST documentation in HBPC remains variable. Previous studies suggest audit and feedback may be insufficient when used alone. This study evaluated whether combining audit and feedback with tailored implementation facilitation can increase and sustain LST documentation rates in HBPC.

Methods

We evaluated the prospective implementation of a longitudinal intervention using retrospective data design with data from the VA Corporate Data Warehouse and HBPC Masterfile between October 2019 and December 2024. Eleven HBPC programs with historically low (< 50%) LST documentation rates participated in a phased intervention consisting of a 6‐month pre‐implementation phase, a 15‐month implementation phase, and a 12‐month sustainability phase. The intervention combined monthly audit and feedback reports with site‐specific implementation facilitation. We used a difference‐in‐differences (DID) analysis to compare changes in monthly site‐level LST documentation rates at intervention sites (Cohorts 1–3) versus non‐intervention sites (Cohort 4). The primary site‐level outcome was the percentage of Veterans with a completed LST template.

Results

The analysis included a total of 140 VA sites, with 11 intervention sites across six VA regions. Intervention sites demonstrated a significant and sustained increase in LST documentation during implementation. The overall average treatment effect was 0.21 (95% CI: 0.144–0.276), corresponding to an average increase over expected trends of 21 percentage points across all intervention cohorts. This effect was maintained throughout the 12‐month sustainability period across all cohorts.

Conclusions

Pairing audit and feedback with implementation facilitation produced a substantial and durable improvement in LST documentation in HBPC settings. These findings support the use of these two complementary, data‐driven implementation strategies to achieve policy goals of goal‐concordant care for seriously ill Veterans.

Summary

• Key points

  • ○
    This multi‐site intervention targeting VA Home Based Primary Care (HBPC) teams demonstrated a significant, sustained increase in life‐sustaining treatment preference documentation.
  • ○
    Improving life‐sustaining treatment preference documentation rates by an average of 21 percentage points suggests that structured implementation can effectively promote goal‐concordant care across diverse settings.
  • ○
    This study contributes to the implementation science literature in geriatrics and palliative care by offering evidence for audit and feedback and implementation facilitation as effective implementation strategies.
• Why does this paper matter?

  • ○
    This manuscript contributes actionable evidence for scaling effective implementation strategies of audit and feedback and implementation facilitation to improve life‐sustaining treatment preference documentation within complex health care systems.

Life sustaining treatment template completion rates. The X‐axis shows the year and month, while the Y‐axis indicates the rate of life sustaining treatment template completion. The black line represents the average observed LST completion rate at comparison sites, which includes all Cohort 4 sites plus all intervention sites prior to initiating the intervention at that site. Vertical reference lines represent the first month of implementation for each cohort. In Cohort 1 (purple), implementation occurred in February 2022; for Cohort 2 (green), implementation occurred in June 2022; and for Cohort 3 (lavender), implementation occurred in November 2022. As with Cohort 4, for each intervention cohort, lines represent the average observed documentation rates.

1. Introduction

Ensuring that seriously ill individuals receive care aligned with their values, goals, and preferences is a cornerstone of high‐quality geriatric and palliative care [1, 2, 3, 4]. For older adults with chronic illness—particularly those enrolled in Home Based Primary Care (HBPC) programs—clear documentation of life‐sustaining treatment (LST) preferences is essential to support patient‐centered decision making, guide surrogate decision makers, and reduce unwanted or burdensome interventions. Despite national efforts to promote advance care planning and goals‐of‐care conversations, documentation of treatment preferences remains inconsistent across healthcare systems, including within the Department of Veterans Affairs (VA) [5, 6].

The Veterans Health Administration (VHA), the largest integrated health care system in the United States, has made sustained investments to normalize documentation of patient preferences through the VA Life‐Sustaining Treatment Decisions Initiative (LSTDI), launched in 2017. LSTDI standardized the process for eliciting, documenting, and honoring Veterans' preferences for life‐sustaining treatments across care settings using a structured LST template embedded in CPRS [1, 6, 7]. LST documentation functions as an actionable medical order recorded in the electronic health record following a goals‐of‐care conversation. LST orders complement, but do not replace, other advance care planning tools and are authoritative within the VA. Because LST orders are not recognized in non‐VA settings, Veterans who receive care in community settings may require state authorized portable orders (e.g., POLST, MOLST).

Early evaluations of LSTDI demonstrated improved documentation and family satisfaction in institutional settings such as VA Community Living Centers [1, 8]. However, uptake has been uneven across VA programs. HBPC programs—serving Veterans who are often medically complex, functionally impaired, and at high risk of hospitalization and death—have shown particularly variable adoption of LST documentation [9]. Although HBPC teams are well positioned to engage Veterans and caregivers in longitudinal, relationship‐based discussions, barriers such as competing clinical demands, variable workflows, and limited implementation support have constrained consistent use of the LST template [9, 10]. As a result, LST documentation rates remain suboptimal in many HBPC programs, even among populations with annual mortality exceeding 30% [11].

This variation has occurred alongside a renewed VA‐wide emphasis on expanding access to palliative care and serious illness services. As the VA increases interdisciplinary palliative care capacity and promotes early integration of serious illness care, reliable documentation of goals and preferences in home‐based settings has become increasingly important to ensure that expanded services remain aligned with Veterans' values.

Evidence from implementation science indicates that multifaceted strategies—particularly those combining audit and feedback with implementation facilitation—can improve uptake and sustainability of evidence‐based practices [12, 13, 14, 15, 16, 17, 18, 19]. Building on prior work conducted through the VA Long‐Term Care Quality Enhancement Research Initiative (QUERI), which identified substantial variation in LST documentation across HBPC programs following initial LSTDI rollout, we evaluated a multi‐site implementation effort, Preferences Elicited and Respected for Seriously Ill Veterans through Enhanced Decision‐Making (PERSIVED), designed to improve LST documentation in HBPC programs with historically low performance [8, 19, 20].

We hypothesized that a structured implementation strategy combining regular audit and feedback with tailored facilitation would significantly increase and sustain rates of LST documentation among Veterans enrolled in HBPC over time.

2. Methods

2.1. Study Design and Data Sources

To evaluate the PERSIVED quality improvement initiative, we conducted a prospective implementation of a longitudinal intervention. This study used data retrospectively collected from October 2019 to December 2024 from the VA Corporate Data Warehouse (CDW), which integrates clinical and administrative data from across VA. Our objective was to increase and sustain rates of LST documentation among HBPC patients by monitoring the impact of our intervention over time. HBPC census data were identified using the HBPC Master File, a national registry maintained by VA Geriatrics and Extended Care, which contains information to identify all HBPC admissions and discharges. Patient characteristics including age, sex, marital status, rurality, race, ethnicity, VA Frailty Index (VAFI), and recent hospitalizations and emergency department (ED) visits for all HBPC patients were extracted from the CDW [21, 22]. The VAFI is an index comprised of 31 elements based on ICD‐10 codes, developed to measure frailty using health records in Veterans [22]. Each element was determined by looking for the occurrence of at least one inpatient or two outpatient encounters with an associated diagnosis code in the year prior to October 2019 within the VA administrative data, VA paid community care, and CMS claims. The number of hospitalizations and ED visits was counted for the year prior to October 2019.

2.2. Site Selection

Eleven out of 136 VA HBPC programs were selected based on low (< 50%) LST documentation rates as of October 2020. These programs represented six Veterans Integrated Service Network (VISN) regions including: New England, Mid‐Atlantic, Southeast, Great Lakes, South Central, and the Rocky Mountain Region. Sites were divided into three groups (cohorts) for the purpose of implementation, and implementation of the intervention occurred using a stepped‐wedge design. However, actual implementation took place at the program level for each of the 11 programs.

2.3. Intervention

The intervention was a structured, multi‐component implementation strategy designed to increase and sustain documentation of LST preferences among Veterans enrolled in VA Home Based Primary Care. The intervention combined monthly audit with feedback and tailored implementation facilitation to support HBPC teams in conducting high‐quality goals‐of‐care conversations and reliably documenting LST decisions in the electronic health record. Audit with feedback involved the systematic collection and reporting of team‐level LST documentation data, with recurring, actionable feedback reports that enabled teams to monitor performance over time, identify gaps, and guide problem‐solving. Together, these components provided clinicians with performance data, training, and practical tools to support behavior change and align care with seriously ill Veterans' goals, values, and preferences.

The PERSIVED intervention consisted of three phases, with HBPC programs going through each phase based on which implementation cohort they were in (Figure 1):

1. Pre‐implementation (Months 0–6): Extensive email and phone interaction with sites involved identifying site champions, assessing barriers and facilitators to adoption of the LSTDI, and process mapping existing workflows. The PERSIVED coaching team used this information to design site‐specific facilitation plans.

2. Implementation (Months 6–21): provision of resources on how to conduct goals of care conversations (written materials and videos) and collaborative problem‐solving sessions informed by a content expert in HBPC practice.

FIGURE 1.

Timeline and PERSIVED intervention components.

Site Managers: Four study team members served as site managers and primary points of contact for the intervention sites. Two team members were assigned to two sites, one team member was assigned to three sites, and one team member was assigned to four intervention sites. Their role was to send feedback reports and rosters, participate in coaching calls, and share resources.

Feedback Reports and Rosters: Sites received monthly feedback reports and individual rosters of patients. These reports were developed with user‐centered design principles to provide visualization of site‐specific LST completion rates [23]. Feedback reports were sent to each site champion and their site leadership.

Coaching Calls: The study team aimed to schedule monthly virtual coaching calls with each intervention site. A site manager and at least one coach were present on each call. Resources were shared during coaching calls, emailed upon request after coaching calls and shared on a Teams channel. Resources included: editable and .pdf versions of the Goals and Preferences Note to Inform the LST Plan template and the LST template; an infographic describing the differences between advance care planning approaches, examples of process maps, and videos demonstrating how to complete an LST template. Resources from the VHA National Center for Ethics in Healthcare included setting health care goals resources, patient education brochure, goals of care conversation guide, a handout on completing the Goals and Preferences to Inform the LST Note, and the LST template.

• 3Sustainability (Months 21–33): Feedback reports continued, but implementation facilitation ended. Sites were expected to sustain improved documentation practices independently.

2.4. Outcomes

Monthly outcomes were tracked for each site, with 63 repeated observations per site between October 2019 and December 2024. The primary outcome measure was site‐level percentage of Veterans with LST documentation who had been admitted to HBPC [20]. Site level outcomes were calculated for all intervention sites (Cohorts 1–3) as well as for all non‐intervention sites (Cohort 4). Sites were classified into intervention groups (Cohorts 1–3) according to the stepped‐wedge design, based on when the intervention was implemented. In the months prior to implementation, sites were categorized as unexposed (not yet received intervention) and in months post‐implementation, sites were categorized as exposed (intervention received). Cohort 4 sites never received the intervention and were categorized as unexposed for all months. Baseline covariates, calculated as of October 2019, were also included and compared across the four cohorts.

2.5. Statistical Analysis

A difference‐in‐differences (DID) model was used to estimate the effect of the intervention on site‐level LST documentation rates [24]. This model allows us to compare the change in LST documentation rates over time at sites with and without the intervention, accounting for multiple measurement times for each site and the staggered implementation of the intervention at different sites over time. The advantage of using this method is that by subtracting changes observed in the comparison group from the changes that occur in the intervention group, DID removes the influence of secular trends that are common to both groups. The assumption is that, conditional on observed site covariates, both groups (the intervention and the comparison group) would follow a similar trajectory. Average treatment effects were estimated for each cohort, each month, and aggregated across all cohorts to show the dynamic effect of the intervention over time. Confidence intervals were calculated using a bootstrap with 500 replications, accounting for clustering due to repeated measurements. Analyses were conducted using R and the Differences‐in‐Differences or “DID” package [24].

3. Results

Across the 11 sites, the number of coaching calls ranged from 6 to 14 calls during the implementation phase, with an average of 11 calls per site. One site only had six coaching calls. The number of champions at each site ranged from two to six, and the rates of attendance by these coaches to coaching calls ranged from 33% to 100%. Coaching calls ranged from 12 to 52 min in length, with an average length of 29 min. Feedback reports and patient rosters were tailored by creating separate reports for each provider at a given site or a report for a specific HBPC team at a given site. Data elements were added at the request of sites, such as including the LST completion rate by provider and separate data for new admissions. Tailoring of the intervention involved changing call frequency to meet champion preferences, engaging subject matter experts on specific calls to augment information provided by the coaches, providing education to HBPC team members who were not serving as champions at times other than regular coaching calls, and engaging additional team members beyond the site champions.

At baseline (October 2019), 136 VA sites were included in the analysis, with 11 sites receiving the intervention across three implementation cohorts (Cohorts 1–3) and 125 sites serving as comparison sites or no intervention (Cohort 4). Table 1 presents patients and site‐level characteristics across these cohorts at baseline. The average age across sites of Veterans admitted to HBPC was consistent across groups, with an overall mean age of 77.8 years (SD 12). Most Veterans were male (95%), with minimal variation across cohorts. Whether Veterans lived in urban or rural locations varied slightly by cohort; Cohort 1 served predominantly urban Veterans (87%), whereas Cohorts 2 and 3 had a higher proportion of Veterans residing in rural areas. Comparison sites (Cohort 4) had a combined enrollment that was 66% urban and 33% rural.

TABLE 1.

Patient and site level characteristics by cohort.

	Total	Cohort 1	Cohort 2	Cohort 3	Cohort 4
Patient characteristics
N (number of patients in each group/cohort)	52,291	789	1535	1254	48,713
Age (mean (SD))	77.8 (12.0)	79.1 (11.5)	78.6 (11.4)	76.6 (12.5)	77.8 (12.0)
Male (N (%))	49,638 (94.9)	751 (95.2)	1455 (94.8)	1189 (94.8)	46,243 (94.9)
Married (N (%))	24,994 (47.8)	322 (40.8)	766 (49.9)	615 (49)	23,291 (47.8)
Highly rural (N (%))	499 (1.0)	1 (0.1)	17 (1.1)	24 (1.9)	457 (0.9)
Rural (N (%))	17,572 (33.6)	105 (13.3)	623 (40.6)	664 (53)	16,180 (33.2)
Urban (N (%))	34,163 (65.3)	683 (86.6)	894 (58.2)	566 (45.1)	32,020 (65.7)
Ethnicity (N (%))
Hispanic or Latino	2393 (4.6)	30 (3.8)	33 (2.1)	11 (0.9)	2319 (4.8)
Not Hispanic or Latino	47,374 (90.6)	712 (90.2)	1450 (94.5)	1169 (93.2)	44,043 (90.4)
Unknown	2524 (4.8)	47 (6.0)	52 (3.4)	74 (5.9)	2351 (4.8)
Race (N (%))
Black or African American	8312 (15.9)	85 (10.8)	251 (16.4)	262 (20.9)	7714 (15.8)
Other/unknown	5277 (10.1)	103 (13.1)	132 (8.6)	111 (8.9)	4931 (10.1)
White	38,702 (74)	601 (76.2)	1152 (75.0)	881 (70.3)	36,068 (74.0)
VAFI (mean (SD))	8.0 (4.1)	8.6 (4.0)	9.0 (4.0)	7.5 (3.8)	8.0 (4.1)
Hospitalizations (mean (SD))	0.5 (1.2)	0.7 (1.4)	0.5 (1.2)	0.2 (0.7)	0.5 (1.3)
ED admissions (mean (SD))	0.5 (1.4)	0.6 (1.3)	0.4 (1.0)	0.2 (0.8)	0.6 (1.4)
Site characteristics
N (number of HBPC programs in each group/cohort)	136	3	4	4	125
Region name (N (%))
Midwest	32 (23.7)	1 (33.3)	0 (0.0)	0 (0.0)	31 (25.0)
Northeast	24 (17.8)	1 (33.3)	1 (25.0)	1 (25.0)	21 (16.9)
South	50 (37.0)	0 (0.0)	3 (75.0)	3 (75.0)	44 (35.5)
West	29 (21.5)	1 (33.3)	0 (0.0)	0 (0.0)	28 (22.6)
Division name (N (%))
East North Central	20 (14.8)	1 (33.3)	0 (0.0)	0 (0.0)	19 (15.3)
East South Central	10 (7.4)	0 (0.0)	0 (0.0)	2 (50.0)	8 (6.5)
Middle Atlantic	16 (11.9)	0 (0.0)	1 (25.0)	0 (0.0)	15 (12.1)
Mountain	13 (9.6)	1 (33.3)	0 (0.0)	0 (0.0)	12 (9.7)
New England	8 (5.9)	1 (33.3)	0 (0.0)	1 (25.0)	6 (4.8)
Pacific	16 (11.9)	0 (0.0)	0 (0.0)	0 (0.0)	16 (12.9)
South Atlantic	25 (18.5)	0 (0.0)	1 (25.0)	0 (0.0)	24 (19.4)
West North Central	12 (8.9)	0 (0.0)	0 (0.0)	0 (0.0)	12 (9.7)
West South Central	15 (11.1)	0 (0.0)	2 (50.0)	1 (25.0)	12 (9.7)
Facility complexity (N (%))
1a	39 (29.3)	2 (66.7)	2 (50.0)	0 (0.0)	35 (28.7)
1b	20 (15.0)	1 (33.3)	0 (0.0)	1 (25.0)	18 (14.8)
1c	28 (21.1)	0 (0.0)	1 (25.0)	0 (0.0)	27 (22.1)
2	20 (15.0)	0 (0.0)	0 (0.0)	0 (0.0)	20 (16.4)
3	26 (19.5)	0 (0.0)	1 (25.0)	3 (75.0)	22 (18.0)

Note: Region: Regions reflect the four Census Bureau‐designated regions; Division: Divisions reflect the nine Census Bureau‐designated divisions; Facility complexity: The facility complexity reflects the level of services provided at a VA facility. Each facility is categorized into one of five complexity level designations (1a, 1b, 1c, 2, 3) with level 1a being the most complex and level 3 being the least complex.

Abbreviations: SD, standard deviation; VAFI, VA Frailty Index.

Visually, cohort‐specific trends (Figure 2) demonstrated a clear upward inflection in LST documentation rates during the intervention period. In Figure 2, the black line represents the average observed LST completion rate at comparison sites, which includes all Cohort 4 sites plus all intervention sites prior to initiating the intervention at that site. Vertical reference lines represent the first month of implementation for each cohort. In Cohort 1 (purple), implementation occurred in February 2022; for Cohort 2 (green), implementation occurred in June 2022; and for Cohort 3 (lavender), implementation occurred in November 2022. As with Cohort 4, for each intervention cohort, lines represent the average observed documentation rates.

FIGURE 2.

Life sustaining treatment template completion rates observed documentation rates. The X‐axis shows the year and month, while the Y‐axis indicates the rate of life sustaining treatment template completion. The black line represents the average observed LST completion rate at comparison sites, which includes all Cohort 4 sites plus all intervention sites prior to initiating the intervention at that site. Vertical reference lines represent the first month of implementation for each cohort. In Cohort 1 (purple), implementation occurred in February 2022; for Cohort 2 (green), implementation occurred in June 2022; and for Cohort 3 (lavender), implementation occurred in November 2022. As with Cohort 4, for each intervention cohort, lines represent the average observed documentation rates.

Figure 3 shows the average treatment effect for each intervention cohort, or the difference in observed changes in documentation (Figure 2) and the expected change assuming no intervention. The expected rate was calculated from the base rate of LST completion in the month prior to implementation. Prior to the intervention (all time points prior to February 2022 when the intervention was initiated for Cohort 1), no significant differences are noted between the three intervention cohorts, supporting the parallel trends assumption. Following implementation, each intervention cohort experienced statistically significant increases in documentation relative to the expected counterfactual.

FIGURE 3.

Difference in differences of life sustaining treatment template completion rates. The average treatment effect for each intervention cohort is shown as the difference in observed changes in documentation and the expected change assuming no intervention. The expected rate was calculated from the base rate of LST completion in the month prior to implementation. The X‐axis shows the year and month, while the Y‐axis indicates the effect on change in life sustaining treatment template completion rates.

Figure 4 aggregates the cohort effects to show the overall effect of implementation over time. Aggregated cohort‐level effects (Figure 4) confirmed a consistent increase in documentation attributable to the intervention. Across all intervention sites, the overall Average Treatment Effect based on this dynamic aggregation was 0.21 (95% CI: 0.144, 0.276). This represents a 21 percentage point greater increase in the rate of LST documentation at intervention sites compared to the counterfactual trend estimated from the comparison cohort and pre‐intervention periods for the intervention cohorts [1, 2, 3]. Dynamic analyses showed that documentation rates increased after implementation and remained elevated over the sustainment period across the three intervention cohorts.

FIGURE 4.

Average effect by length of exposure. Aggregate cohort effect to show the overall effect of implementation over time. The X‐axis shows months after implementation, while the Y‐axis indicates the effect on change in life sustaining treatment template completion rates.

4. Discussion

In this multi‐site evaluation, the results suggest that an intervention combining monthly audit and feedback with tailored implementation facilitation led to a significant, sustained increase in the rate of LST documentation compared to no implementation support in comparison sites. The average treatment effect of a 21 percentage increase in LST documentation, in comparison with facilities that did not receive the intervention, suggests a clinically meaningful impact of the intervention, which was sustained over the 12‐month post‐implementation period.

In a similar study, which implemented audit and feedback alone without facilitation among 13 HBPC sites, no significant difference was observed in intervention compared with matched comparison sites (LST rates rose from ~6% to 42% in both groups) [9]. In that study, the fact that documentation rates increased similarly across all sites suggests that national rollout efforts may have driven changes irrespective of the audit and feedback intervention and that additional implementation strategies beyond audit and feedback may be needed to affect a significant change. Our results suggest that the addition of implementation facilitation to audit and feedback accelerated and sustained gains in documentation. Engaging in facilitation activities during coach calls, such as goal setting, problem solving, tailoring the intervention to specific sites, and refining workflow and processes, augmented audit and feedback and enabled sites to realize measurable change.

These findings align with a recent Cochrane review of the effects of audit and feedback on professional practice [12]. Our study incorporated many of the characteristics associated with effective audit and feedback interventions, for example, providing feedback on important performance metrics where health professionals have substantial room for improvement. In this study we included sites if LST completion rates were less than 50%. Further, at the request of individual sites, reports were modified to include measures of individual practice (i.e., data separated out by HBPC provider), making them more actionable, rather than providing feedback at the site level only.

We involved local champions to receive the feedback and communicate that feedback with their colleagues. During calls with each team, our feedback intervention included asking champions if they had reviewed the report, soliciting their thoughts about the most recent feedback report and patient roster, exploring any questions or concerns they had about the data, and asking if they had shared the report with their team or, if not, if they had plans to do so. The feedback reports compared performance to peers, another component that enhances effectiveness. Finally, as identified in the 2025 review, our intervention involved joint creation—shared between the coaching team and the local team—of actionable plans with tailored advice for improvement during coaching calls [12]. This multi‐component approach may explain the strong and sustained effects observed in this study. Importantly, qualitative data indicate that feedback alone without these additional co‐interventions may be demoralizing if recipients perceive poor performance without guidance for improvement [9]. Facilitation helps overcome this barrier by contextualizing data, building self‐efficacy among clinicians and increasing engagement [25, 26, 27].

Our results suggest that a multi‐component, tailored intervention, particularly one that acknowledges and supports the HBPC workforce and dispersed workflow realities, is essential for sustained practice change. This aligns with findings from several studies, including a recent scoping review, recommending use of practice‐focused resources to support implementation of interventions in the settings where people live [28, 29, 30]. In the scoping review, the authors recommend:

1. resources to support evidence‐informed practice, such as LST documentation;

2. resources that clearly define their target audience and tailor communication to this audience, which we accomplished during our coaching preparation and calls;

3. resources that draw on evidence from a range of sources such as the feedback reports we used in our study;

4. resources with practical implementation strategies like having an HBPC provider involved in coaching; and

5. resource content adapted to different contexts, such as our tailored coaching for each team.

Key strengths of this study include the rigorous difference‐in‐differences design with a comparison group, geographically diverse sample, and evaluation of both short‐term and sustained effects. This approach addresses potential confounding by leveraging temporal trends and accounts for secular changes in documentation practices. The intervention was theoretically grounded and informed by user‐centered design principles to enhance its acceptability and feasibility across sites.

This study also has several limitations. First, the study did not use a randomized design and HBPC programs were not randomly assigned to cohorts, or steps in the stepped wedge design. Although we adjusted for baseline trends and included comparison sites, there may still be unmeasured confounding. Second, while sites were geographically dispersed and included urban and rural locations, we had a small number of intervention sites, potentially limiting generalizability and power. However, we focused on those sites that were performing poorly at baseline, so that our intervention was delivered to sites that needed the most support. Third, we focused on LST preference documentation as the primary outcome and did not assess whether documentation translated into meaningful caregiver‐centered outcomes such as care quality, concordance of care preferences with treatment or satisfaction. However, documentation is an essential implementation/process outcome required to enable goal‐concordant care. Fourth, one site continued to receive monthly feedback reports but elected to stop coaching during month seven of 21. This site was retained in the analysis although it was not exposed to the intervention for the same duration or intensity as the other sites. Fifth, generalizability is limited given that this was a predominantly male population who received care in a health care system with a well‐integrated electronic health record.

In summary, while national efforts such as the LSTDI have improved documentation rates, our findings suggest that audit and feedback augmented by implementation facilitation may accelerate and sustain improvement in HBPC settings. The growth in care delivery in home‐based settings makes this a compelling issue. The minimum amount of facilitation needed to achieve benefit should be explored given that facilitation is resource‐intensive. Such data would provide actionable evidence for health systems seeking to close implementation gaps in the care of seriously ill patients while demonstrating the value of investing additional resources in contextualized, hands‐on support to implementing practice change.

Author Contributions

C.R.L. wrote the main manuscript text. P.L. conducted the analyses and prepared the tables and figures. A.S. made substantial contributions to the analytic plan and conception of the work. All authors revised the manuscript for intellectual content. All authors read and approved the final manuscript.

Funding

This project was supported through a grant from the Veterans Health Administration (VHA) Health Services Research and Development Service Quality Enhancement Research Initiative (QUE 20‐015).

Disclosure

During the preparation of this work, Dr. Levy used Chat GPT to improve the readability of the manuscript. After using this tool, Dr. Levy reviewed and edited the content and takes full responsibility for the content of the publication. The PERSIVED team retained full independence in the conduct of this program and the views of this publication are the authors' own and do not necessarily reflect those of the Department of Veterans Affairs.

Conflicts of Interest

The authors declare no conflicts of interest.