State POLST Program Maturity Status and Dying in the Nursing Home or Hospice in the United States: An Event Time Study

Abstract

Objectives:

State Physician Orders for Life-Sustaining Treatment (POLST) programs ensure orders for treatment are documented for the provision of goal-concordant care at the end of life. The national POLST organization tracks the maturity stages of state programs and categorizes them as developed (beginning of use) and endorsed (benchmarks and standardized use). Examining the association between POLST program maturation and the likelihood of dying in place can help clarify how implementation stage may affect end-of-life care and place of death. This study examined the association between the state POLST program maturity stage with place of death of nursing home (NH) residents (eg, nursing home or hospice).

Design:

An event time study.

Setting and Participants:

This study included a 10% random sample of NH residents aged 65 and older who died between 2012 and 2018 in the United States.

Methods:

Data for NH residents were merged with data on each state’s POLST development and endorsement year. Place and date of death were identified using the Minimum Data Set 3.0 and National Vital Statistics. A linear probability model assessed the likelihood of dying in a NH or hospice vs elsewhere before and after POLST implementation.

Results:

Among 225,149 NH residents, 74.3% had Alzheimer’s disease or dementia, 67% were women, and 82.2% were White. Following POLST program development, the probability of dying in a NH or hospice increased by 1.6 percentage points (pp) in the first year (95% CI, 0.006–0.027; P = .003) and by 5.6 pp after 5 years (95% CI, 0.041–0.070). A sustained increase in likelihood of dying in place was also observed after POLST endorsement.

Conclusions and Implications:

POLST program maturity is associated with a higher likelihood of NH residents dying in place or in hospice, potentially supporting goal-concordant care and reducing burdensome hospital transfers near the end of life.

Planning for end-of-life care using advance directives in the nursing home (NH) setting is often fragmented and unstandardized across the United States.¹ In response, state Physician Orders for Life-Sustaining Treatment Program (POLST) programs were created to assist in facilitating decision-making processes at the end of life to ensure preferences for treatment are stated before death, which is intended to be a collaborative process involving caregivers as well. ² POLST orders delineate the preferences and choices of seriously ill patients regarding code status, life-sustaining treatments, and other end-of-life decisions such as feeding tubes. ³ Of note, there are a number of different names for state POLST programs (eg, POST, MOST, MOLST, and TPOPP); we use POLST in this report.

POLST programs have been implemented heterogeneously across US states, with substantial variation in timing, scope, and degree of program maturity. Some states have long-standing, formally endorsed POLST programs integrated into clinical practice and others have more limited or emerging implementation. This variation reflects differences in state policies, clinical uptake, and infrastructure supporting serious illness and end-of-life care. Understanding this evolving landscape is essential for examining how POLST program maturity stage may influence end-of-life decision making and outcomes (eg, place of death) for NH residents. ^4–6

State programs have been categorized into 4 maturation stages by the national POLST organization based on the level of implementation. These stages are developing, endorsed, matured, and non-conforming. In the developing stage, a state’s program is in the initial stages of development and implementation. ⁷ A POLST program achieves an endorsed stage when there is statewide standardized use and implementation benchmarks such as a single-use form, continuing education, and quality assurance. In the mature stage, a state has incorporated the POLST program as a widely implemented standard of care for the seriously ill. A nonconforming program means that it does not meet the standards for endorsement by the national POLST organization. By 2020, few states had achieved the mature stage and most remained in the development and endorsed stages.

Since the inception of POLST, despite variable implementation and heterogeneity of adoption across the United States, programs have largely been reported to be linked with improved quality of care and completion of advance directives. ⁵ However, ongoing research demonstrates that although POLST programs have favorable outcomes, many NH residents still do not receive goal-concordant care at the end of life and face burdensome and costly transitions in the context of acute and worsening illness or exacerbations in the NH.^{6 ,8} Implementation and maturation of POLST programs is critically important for NH residents. POLST forms document their health care preferences, become readily accessible, and can be consistently honored across care settings. Residents in long-term care facilities often have heightened medical complexity and multiple chronic conditions^{9 ,10} or are nearing the end of life, placing them at greater risk of acute hospitalizations that require immediate end-of-life decision making. POLST orders provide clear and directive guidance for timely and appropriate medical interventions aligned with the resident’s goals of care.

Longitudinal analyses of the association between maturation of state POLST programs and association with place of death of NH residents have not been conducted. Deeper understanding of the impact on this key end-of-life outcome (place of death) has potential to inform implementation, program, and policy evaluation efforts across stages and facilities. The objective of this study was to examine the association between POLST maturity status (developing or endorsed) and place of death (NH or hospice) of NH residents over time using an event time study approach.

Methods

Study Sample

The study population included NH residents older than 64 years who died sometime between January 2012 and December 2018. Participants were identified using the Minimum Data Set 3.0 (MDS), a standardized, comprehensive assessment tool used for all residents at admission and at routine intervals in federally licensed NHs in the United States. NH-resident characteristics (eg, demographics, clinical) were obtained from the last quarterly MDS assessment available within 1 year of death for each resident. The MDS data were merged with Vital Statistics Data, which include dates for all deaths during the study period. The MDS assessments and dates of death were then merged with the longitudinal (ie, between 1990 and 2020) national POLST organization data on the maturation stage of each state program per year. As part of the analytic process, we conducted a scan of the national POLST website to identify and classify the timing, maturity, and endorsement stage of POLST programs across US states and used these classifications to define state-level POLST exposure variables included in the analysis.

This study was approved by the institutional review boards at Columbia University and RAND Corporation. The study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) ¹¹ reporting guidelines.

Exposure

NH residents were considered exposed to POLST if they were alive during the year before the development or endorsement of POLST within the state where they ultimately died.

Outcomes

The primary outcome was the location of death—specifically, whether the resident died in place (eg, in an NH or hospice) vs elsewhere (eg, in a hospital)—within a given calendar year and state.

Covariates

We created individual-level indicator variables for age at death in 1-year increments (65 as a reference). We also created indicators for whether the resident in the last MDS quarterly assessment had an active diagnosis of pneumonia, diabetes, end-stage renal disease, cancer, stroke, congestive heart failure, Alzheimer’s disease or related dementia, or chronic obstructive pulmonary disease. Using the MDS, we also created indicators for race, sex, and marital status. We include individual-level covariates from the last available assessment (or the last nonmissing data from one of the last assessments within 1 year of death) to improve precision of our treatment effect estimates and account for compositional differences across states and over time.

Statistical Analysis

First, we counted the number of states in each POLST maturity and endorsement category over time to describe national patterns of POLST implementation. Next, we performed descriptive analyses for characteristics and outcomes of residents. We reported unadjusted means with standard deviations for continuous variables and proportions for categorical variables. We then estimated the association between probability of dying in place (eg, in an NH or hospice) or elsewhere (eg, hospital) and POLST maturity at the NH-resident level. We used an event study approach to first obtain the magnitude of this association for each year after POLST was developed relative to the year before its development. We estimated our regression specification using the linear probability model. Our outcome was a linear function of yearly binary indicators measuring time since POLST was developed in each state. These indicators thus identified changes in outcomes in yearly increments relative to the probability of dying in place (vs elsewhere) 1 year before POLST development (the reference year, or t-1). We adjusted these regressions by including state indicators to adjust for time-invariant unobserved factors associated with both POLST and probability of dying in place; calendar year and month of death indicators to adjust for nationwide secular trends and seasonality in probability of dying in an NH or hospice (year 2012 and month January were omitted as reference); age categories (65 years was the reference); and a list of individual characteristics such as race, sex, marital status, and multiple chronic conditions. These regressions were limited to states not yet endorsing POLST.

We then assessed the association between the probability of dying in place following the POLST endorsement using a similar model. Because POLST endorsement always occurred after POLST development, these regressions were limited to states that already had developed POLST. We expected to observe a change in trend in the probability of dying in place, potentially driven by years of POLST development, even before its endorsement. Thus, we sought to identify the difference between the estimated outcomes and the counterfactual trend (ie, what would have happened if the states had not endorsed POLST). These analyses were intended to evaluate whether exposure to POLST is associated with end-of-life care that aligns with an intended outcome, increasing the likelihood that individuals die in place within an NH or hospice, rather than elsewhere. Specifically, we controlled for a linear trend in probability of NH or hospice death before POLST endorsement, and yearly indicators measuring for time after the POLST endorsement. To identify the counterfactual trends, we projected the observed linear trend into the postintervention period. Therefore, the coefficients on yearly indicators measured the change in outcomes post-POLST endorsement relative to any preexisting linear trend in probability of dying in place. The decision to model a linear trend was based on the patterns observed in our data.

We computed 95% CIs adjusted for clustering within states. We used Stata/MP statistical software, version 16.0 (StataCorp LLC) for analyses. Two-sided P values were deemed statistically significant at <.05.

Results

Study Sample

Our findings represent approximately 225,149 NH residents (Table 1). Two-thirds (67%) were women and most were White (82.2%). Common among NH residents with complex care needs, multiple chronic conditions were highly prevalent. Common conditions included Alzheimer’s disease or Alzheimer’s disease and related dementias (AD/ADRD) (73.3%), diabetes (32.8%), congestive heart failure (30.1%), coronary artery disease (24.7%), end-stage renal disease (16.8%), chronic obstructive pulmonary disease (26.2%), and stroke (15%).

Table 1.

Descriptive Statistics for NH Decedents Between 2012 and 2018 With at Least 1 NH Quarterly Assessment Before Death, n = 225,149

	%	SD
Died in NH/hospice	76	0.42
Age	85.7	8.44
Female	67	0.47
American Indian or Alaskan Native	0.3	0.057
Asian	2	0.12
Black or African American	9	0.29
Hispanic or Latino	4	0.19
Native Hawaiian/Pacific Islander	0.3	0.05
White	82	0.38
Diabetes	33	0.47
Pneumonia	7	0.25
Heart failure	30	0.46
Stroke	15	0.36
End-stage renal disease	17	0.37
Coronary artery disease	25	0.43
Alzheimer’s disease	19	0.40
Dementia	54	0.50
Chronic obstructive pulmonary disease	26	0.44
Cancer	9	0.29
ADL score	20.94	5.14
ADL = 28	0.12	0.33

ADL, activities of daily living.

10% random sample. Information was pulled from the latest assessment for which the person had available/nonmissing data (not the latest quarterly assessment). Summary statistics based on years 2012–2018.

POLST Maturity Status Across the United States

By 2010, 33 states had developed POLST programs and an additional 14 developed POLST by 2019. Figure 1A and B shows the variation in POLST development and endorsement across states over time. POLST endorsements began in 2004. By 2010, 9 states had endorsed POLST programs, and by 2020, another 17 states had endorsed programs.

Fig. 1.

(A) Distribution of states with developed programs over time (1990–2020). (B) Distribution of states with endorsed programs over time (1990–2020).

Place of Death of NH Residents

Figure 2 presents the year-to-year unadjusted share of people dying in the NH or hospice over the period between 2012 and 2018 on average across all states. In 2012, 74.7% of residents died in the NH or hospice. By 2018, 78% of NH residents died in the NH or hospice.

Fig. 2.

Percentage of NH residents who died in place over time (2012–2018).

We describe adjusted changes in the probability of dying in place (eg, NH or hospice) following POLST development or following endorsement in Figure 3A and B, respectively. There were no significant changes in probability of dying in place (vs elsewhere) before POLST development (periods from t-2 to t-4) relative to the reference period in t-1 (see Figure 3A). Compared with the probability of dying in place in the year before POLST development (t-1), the probability of dying in place was higher by 5.6 percentage points (pp) (95% CI, 4.1–7.0; P = .000) after 5 years (t + 4). Relative to pre-POLST development, the prevalence of dying in place increased over time. Specifically, it was about 7.7 pp (95% CI, 6.2–9.3; P = .000) or about 10% higher compared with what was expected in the absence of POLST development. Age, cancer, and AD/ADRD increased the probability of dying in place, whereas having heart failure, coronary artery disease, diabetes, or pneumonia, and race other than White decreased it, although these covariates serve primarily as controls rather than as focal predictors. See Supplementary Table 1 for regression estimates supporting Figure 3A and B.

Fig. 3.

(A) Change in the probability of dying in the NH or hospice relative to the year before POLST was developed. (B) Change in the probability of dying in the NH or hospice relative to the year before POLST was endorsed (among states that developed POLST). Markers represent the estimated association at the event time (ie, for each year relative to the POLST development [A] or endorsement [B] year); whiskers, 95% CIs; dashed lines, estimated linear association between probability of dying in an NH or hospice and event time in the absence of POLST development or endorsement, indicating a counterfactual trend in the outcome. (A) Limited to states not yet endorsing POLST. (B) Limited to states that already developed POLST.

In states with POLST programs, the probability of dying in the NH or hospice increased steadily with programs that had reached endorsement and mature status over time, although deviations from this linear trend were not statistically significant. Figure 3B shows that by year 5 since POLST endorsement, prevalence of NH or hospice death increased by approximately 2.4 pp on average (95% CI, −0.9 to 5.7; P = .141) compared with the expected outcome in the absence of POLST, although this difference was not statistically significant. These changes stabilized thereafter, with no significant deviations from the predicted pre-endorsement trend.

NH or hospice death rates in POLST-adopting states increased steadily over time, both before and after policy endorsement (Figure 3B). Postendorsement estimates diverge modestly from the predicted pre-trend, reaching 2.4 pp higher by year 5 (95% CI, −0.9 to 5.7; P = .141) and plateauing at approximately 2.6 to 0.3 pp thereafter. These deviations were not statistically significant, indicating we cannot rule out continuation of preexisting trends. Regression estimates are provided in Supplementary Table 1.

Discussion

To our knowledge, this is the first study to examine the longitudinal association of state POLST programs and the place of death for NH residents using an event time study approach. This study elucidates the potentially positive impact of POLST programs on place of death of NH residents. The findings also suggest that with continued implementation and greater degree of maturation, POLST programs can streamline advance directives in this population and help lessen unnecessary hospitalization and use of aggressive care at the end of life that is often misaligned with the preferences of patients and families. These findings also fill a gap in the literature as a recent systematic review of POLST research revealed only 4 studies examined place of death and none focused solely on NH residents and their place of death.⁵

Although there was a greater likelihood of NH or hospice death in the developing stage, our finding that endorsement was not significantly associated with dying in place or hospice suggests that endorsement may be an insufficient proxy for program maturity or effectiveness. One possible explanation is that endorsement represents a formal recognition of the POLST program by the state but does not necessarily reflect the extent of implementation, provider training, or uptake within NHs. It is also possible that there is a lag between endorsement and meaningful changes in practice, during which time the infrastructure and systems needed to support POLST use are still developing. This could explain why endorsement alone did not correspond to a measurable increase in alignment with the intended outcome. These findings suggest that program endorsement may be necessary but not sufficient for realizing the benefits of POLST, and that other factors, such as sustained education, statewide registry systems, and integration into electronic health records, may be critical to program impact.

Overall, these findings suggest that the maturity of state-level POLST programs is associated with a meaningful increase in the likelihood that NH residents die in an NH or hospice, particularly following program development rather than formal endorsement alone. After POLST development, the probability of dying in place increased progressively over time, reaching a clinically meaningful increase relative to expected pre-policy trends, indicating that more established POLST infrastructure may facilitate care trajectories that avoid unwanted transfers near the end of life.

Our findings show that state POLST endorsement alone may formalize existing practices rather than produce immediate shifts in outcomes. The results indicate that earlier phases of POLST program maturation, when policies, clinical workflows, and clinician uptake are actively evolving, appear to exert the strongest influence on place of death. This pattern underscores that the impact of POLST programs on NH or hospice death depends not only on policy presence, but on the depth of implementation and integration into routine care.

On a practical level, there is potential that continued proliferation and maturation of programs within states can improve impact on patient care and decision making, which would require further investigation. For example, NH residents who are transitioned to hospice have a lower likelihood of hospitalization at the end of life, but the role of POLST in priming or guiding a conversation in relation to a transition to hospice is unclear.¹² The literature reporting on outcomes of POLST programs has mostly included cross-sectional observations that highlight favorable quality of care and POLST outcomes across states.⁵ However, none have been conducted to examine outcomes of the program over time on place of death, transition to hospice, or prevention of hospitalizations and unwanted emergency department visits. Also, given the heterogeneity of implementation of POLST programs, it remains difficult to ascertain the nationwide effects of the program and limits the development of responsive policy efforts.¹³

Most POLST-focused research has included patients who largely identify as White, which is consistent in this study. Increases in POLST use and do-not-resuscitate (DNR) status have largely been documented for White NH residents. Black and Hispanic NH residents have been found to have a greater preference of cardiopulmonary resuscitation (CPR). More analyses reflecting POLST decisions and DNR vs CPR status preferences of Black and Hispanic NH residents are needed.¹⁴ Additional research is also needed to examine the POLST use and outcomes for racially and ethnically diverse NH residents and those receiving care in marginalized or lower-resourced settings. Area deprivation, social determinants of health, and cultural differences in end-of-life care preferences, which are not reflected in standard POLST orders, may influence care decisions.

Future research should consider POLST implementation and facilitation at the facility level. Understanding the characteristics of facilities (eg, staffing, operations, POLST champions) that are more likely to lead to consistent use can shed light on this issue. Further, operationalizing rates of completion should include recall and decision support quality outside of times when emergencies necessitate real-time decision making about end-of-life care.⁸

Finally, a key question about the efficacy of POLST programs remains in that it is unclear if such programs and associated completion of documented orders that outline life-sustaining treatments actually result in goal-concordant care at the end of life for seriously ill NH residents. Given concordance with preferences is often a dynamic phenomenon for many patients and families, understanding how POLST orders can serve as conversation starters or facilitators may be a more patient- and family-centered approach to end-of-life counseling and decision-making.¹⁵

Limitations

This study has limitations. These findings are associations and trends but cannot be interpreted as causal. We were not able to identify the exact mechanisms underlying this study’s findings. Although POLST development or endorsement are important factors associated with dying in place for NH residents and although we accounted for secular trends, other factors happening around the same as development or endorsement beyond what we controlled for could influence our results. Furthermore, this study is limited to NH residents and does not include community-dwelling persons with serious illness.

Conclusions and Implications

State POLST programs that have time to develop have potential to improve end-of-life care quality and decision making in the NH setting. As these programs continue to grow, considerations such as communication quality, goal concordance, and hospice transitions as well as reduction of unwanted hospitalizations need high-quality measurement. With greater study, POLST programs and policies can be helpful in ensuring NH residents can die in place or in hospice should they desire, potentially avoiding discordant and aggressive care transitions that may not be aligned with the treatment preferences of NH residents and their caregivers.

Supplementary Material

Supplementary data related to this article can be found online at https://doi.org/10.1016/j.jamda.2026.106134.