An Integrative Review of the State of POLST Science: What Do We Know and Where Do We Go?

Abstract

Objectives:

POLST is widely used in the care of seriously ill patients to document decisions made during advance care planning (ACP) conversations as actionable medical orders. We conducted an integrative review of existing research to better understand associations between POLST use and key ACP outcomes as well as to identify directions for future research.

Design:

Integrative Review

Setting and Participants:

Not applicable.

Methods:

We queried PubMed and CINAHL databases using names of POLST programs to identify research on POLST. We abstracted study information and assessed study design quality. Study outcomes were categorized using the international ACP Outcomes Framework: Process, Action, Quality of Care, Health Status, and Healthcare Utilization.

Results:

Out of 94 POLST studies identified, 38 (40%) had at least a moderate level of study design quality and 15 (16%) included comparisons between POLST vs. non-POLST patient groups. There was a significant difference between groups for 40/70 (57%) ACP outcomes. The highest proportion of significant outcomes were in Quality of Care (15/19 or 79%). In subdomain analyses of Quality of Care, POLST use was significantly associated with concordance between treatment and documentation (14/18 or 78%) and preferences concordant with documentation (1/1 or 100%). The Action outcome domain had the second highest positive rate among outcome domains; 9 of 12 (75%) Action outcomes were significant. Healthcare Utilization outcomes were the most frequently assessed and approximately half (16/35 or 46%) were significant. Health Status outcomes were not significant (0/4 or 0%), and no Process outcomes were identified.

Conclusions and Implications:

Findings of this review indicate that POLST use is significantly associated with a Quality of Care and Action outcomes, albeit in nonrandomized studies. Future research on POLST should focus on prospective mixed methods studies and high-quality pragmatic trials that assess a broad range of person and health system level outcomes.

Brief Summary/Tweet:

A majority of advance care planning outcomes including quality of care and action outcomes were positively associated with use of POLST in comparison to usual care. Future POLST research should focus on high quality implementation studies.

1. Background

Care planning helps patients and families prepare for communication and informed medical decision-making to support both advance and in-the-moment decisions.¹ Advance care planning (ACP) discussions involve exploring what quality of life means for the patient and are influenced by the patient’s readiness to engage, preferences for decisional control, prognostic awareness, and perspectives about acceptable tradeoffs. The nature of these decisions varies depending on the patient’s life stage (e.g., healthy vs. seriously ill vs end of life) and setting of care.¹ Patients with serious illness or end-stage disease may be asked to make decisions including identifying a surrogate decision-maker, specifying current goals of care, and stating preferences regarding the use of life-sustaining interventions such as cardiopulmonary resuscitation, hospitalization, and intubation. These treatment decisions become increasingly preference sensitive as the burdens of available interventions increase and the likelihood of benefit decreases. Given that a significant proportion of persons near the end of life lose decisional capacity,^2–4 documentation is important to ensure these decisions are recorded and available to health care providers as patients move across settings of care.

POLST was developed in the mid-1990s to help ensure goal concordant care for persons with life-limiting illness at risk of life-threatening events by documenting patient treatment preferences as actionable, portable medical orders.⁵ It is designed to be completed following an ACP conversation and shared decision-making between a health care provider and patient or their surrogate decision maker. Although POLST used to be an acronym for “physician orders for life-sustaining treatment,” that is no longer the case. In 2018, the National POLST Paradigm transitioned to using “POLST” as a standalone word, defined as a “portable medical order,” moving away from its original acronym form.⁶

There have been three prior systematic reviews of POLST. The first published review of POLST research on use in the clinical setting found evidence that POLST supported person-centered care by permitting a wide degree of individualization and that it was viewed positively by a range of health care providers despite numerous practical challenges.⁷ A second review found concordance between POLST treatments and orders⁸ and a third found associations between POLST orders limiting treatments and reduced treatment intensity.⁹ Each of these reviews focused on specific outcomes related to the original goals of POLST, but many studies included in these reviews lacked comparison groups of non-POLST users, making it difficult to assess the effect of POLST on outcomes.

In the last few years, the science of ACP has progressed suggesting a broader review of outcomes may be needed. A 2017 international Delphi panel convened experts in palliative care and created an ACP Outcomes Framework with the following domains: Process (e.g., readiness), Action (e.g., communication and documentation), Quality of Care (e.g., goal concordant care, satisfaction), Health Status (e.g., quality of life, depression) and Healthcare Utilization (e.g., hospitalization).¹⁰ A 2021 scoping review of 65 high-quality clinical trials focused on ACP evaluated outcomes of trials using this framework. Findings suggested high rates of statistically significant differences in outcomes for the Process (72%) and Action (86%) domains, with mixed results for Quality of Care domain (53%), and lower rates of positive outcomes for Health Status (37%) and Healthcare Utilization (42%).¹¹ However, none of the trials included in the review focused on use of POLST, making it difficult to understand of how POLST use effects the broader array of ACP domains.

We conducted this integrative review to systematically analyze and synthesize POLST research.¹² The purpose of this review was to summarize the current state of POLST research and to describe the effects of POLST on ACP outcome domains using the standardized ACP outcomes domain framework.

2. Methods

This review follows Whittemore and Knafl’s procedure for integrative reviews, which includes: (1) problem identification (explained above), (2) literature search, (3) data evaluation, (4) data analysis, and (5) presentation. Integrative reviews enable understanding of a larger body of literature through systematic analysis and synthesis of independent studies. This approach helps increase the generalizability of findings and identification topics for future research, while following well-defined and systematic methods to protect against investigator bias and generate accurate findings.¹²

2.1. Literature Search

We created a list of search terms to identify research on POLST using National POLST’s list of POLST program names;¹³ these search terms included:

“Physician Orders for Life-Sustaining Treatment”) OR “POLST” OR (“Medical Orders for Scope of Treatment”) OR (“Medical Orders for Life-Sustaining Treatment”) OR (“Physician Orders for Scope of Treatment”) OR (“Iowa Physician Orders for Scope of Treatment”) OR (“Louisiana Physician Order for Scope of Treatment”) OR (“Provider Orders for Life Sustaining Treatment”) OR (“Physician Orders for Sustaining Treatment”) OR (“Transportable Physician Orders for Patient Preferences”) OR (“Practitioner Orders for Life-Sustaining Treatment”) OR (“Portable Orders for Life-Sustaining Treatment”) OR (“Pennsylvania Orders for Life-Sustaining Treatment”) OR (“Life with Dignity Order”) OR (“Clinician Orders for Life-sustaining Treatment”) OR (“Providers Orders for Life-Sustaining Treatment”

All state specific names for POLST were included. Only the POLST acronym was included due to its prevalence of use within the field. We originally ran queries using the search terms in PubMed and CINAHL on December 9, 2020 and updated our search results on August 3, 2021 and May 20, 2022. One article was additionally retrieved by hand-searching; this article was written in 1996 before the POLST name or acronym had been created. We imported all retrieved citations into Covidence, a web-based collaboration software platform that streamlines the production of systematic and other literature reviews.¹⁴

2.2. Manuscript Screening

The authors used Covidence to deduplicate citations and conduct all steps of the screening, eligibility, and information extraction phases.¹⁴ Each title and abstract were screened for inclusion by two reviewers (SH and either EU or MF). Manuscripts were included if they reported research on POLST. Next, two of four reviewers (EU, MF, RL, SH) reviewed each full-text article for inclusion, excluding those that did not address POLST directly; reported a quality improvement study, pilot or feasibility study, or literature review; were conference abstracts, editorials, or news articles; solely addressed the presence of POLST (e.g., the number of POLST forms) as the outcome of the study; reported non-human subjects’ research; was published in a language other than English; lacked information about study design; and was a brief description of a larger study that was already included in the review. Figure 1 depicts the flow of information through the phases of the review.¹⁵ For the outcomes extraction, we reviewed the remaining articles to identify studies which examined the effects of POLST in studies with patients who used POLST and a comparison group of those who did not use POLST.

Figure 1.

PRISMA flow diagram. Adapted from PRISMA extension for scoping reviews (PRISMA-ScR): checklist and explanation¹⁴

2.3. Data Extraction and Evaluation

One of four reviewers (EU, MF, RL, SH) extracted the following information from each of the included manuscripts: Study location, purpose, design, setting, number of sites, sample, number of participants, data collection dates, method, and summary of findings. To ensure accuracy, the extracted information was reviewed by a second reviewer (EA, EF, HL, AHM, NW), all of whom were members of the National POLST Research and Quality Assurance Committee. If a discrepancy was found between extracted information and secondary review, changes were made on a case-by-case basis (EU, TM) and unresolved discrepancies were resolved by expert review (SH). Studies with both a group of patients who used POLST (i.e. the “intervention”) and a group of patients who did not (i.e., the “control”), were further examined by two reviewers (EU, SH) to identify outcomes, which were extracted and categorized using the Advance Care Planning Outcomes Framework.^7,8 Each manuscript was also reviewed by at least two reviewers to determine if race or ethnicity was measured among POLST users and if any differences were found between racial or ethnic groups.

The heterogeneity of study designs, approaches, and level of description (i.e., a known limitation of integrative and other reviews) precluded evaluation using a harmonized approach.¹⁵ As such, we classified all studies according to the JAMA-endorsed Quality Rating Scheme for Studies and Other Evidence.^16,17 In this scheme, studies are categorized based on the level of certainty one can have in their findings based on the methodological approach; i.e., randomized controlled trials or systematic reviews with meta-analysis represent the highest level of evidence (Level 1), followed by controlled trials without randomization and prospective cohort trials (Level 2), case control and retrospective cohort studies (Level 3), case series cross-sectional studies (Level 4), and opinion pieces and case reports represent the lowest level of evidence (Level 5).

2.4. Data Analysis

Descriptive statistics were used to describe study characteristics (location, setting, sample, and data collection method).

Following the methodology described by McMahan et al.,¹⁰ we conducted an analysis of included manuscripts to determine the effects of POLST use among the subset of articles with patients who used and did not use POLST to permit comparisons. Primary and secondary outcomes of included studies were categorized according to the standardized ACP Outcomes Framework, including Process, Action, Quality of Care, Health Status, and Healthcare Utilization.^10,11 Outcomes between POLST users and non-POLST users were identified as “positive” if there was a statistically significant difference (p<0.05) between POLST and non-POLST participants. When some aspect of healthcare utilization (e.g., transfer to ICU) could be assessed for concordance with a specific POLST order (e.g., comfort measures), that outcome was classified as a Quality of Care outcome (i.e., did the treatment make sense given the order?). Discrepancies in coding of outcomes between reviewers were resolved through deliberation.

3. Results

3.1. Study Characteristics

Ninety-four manuscripts were identified in this integrative review. The first article was published in 1996 and although research publication was sporadic in early years, the number of publications has steadily increased from 2009 to 2019, with a slight decline in 2020 and 2021 (the most recent search was conducted through June 2022; see Figure 2). Complete citations and data extraction for all included manuscripts are provided in the Supplementary Materials (see Tables S1 and S2).

Figure 2.

Number of POLST research publications from January 1996 to June 2022.

Figure 3 demonstrates the number of POLST publications by study location, which states use POLST, and which states’ POLST programs were endorsed as meeting National POLST program standards. Nearly a third of research studies were conducted in Oregon (n= 27). West Virginia (n= 9), Indiana, New York, Wisconsin (n= 7 each), and California (n= 7) were the next most representative states in our included sample of literature. Only two studies were nationally representative of POLST use across the U.S. (not counted in individual state use counts). Eleven studies were conducted in countries other than the U.S.

Figure 3.

Number of POLST publications by study location and POLST use status in 2022 according to National POLST. Publications are counted more than once if the study sample included multiple states or countries.

The most common settings of the included POLST studies were nursing facilities (n= 34), hospitals or individual units in hospitals (n= 26), and through state registries (n=10). The most common participants in studies include patients from multiple settings (n= 41), nursing facility residents (n=18), and health care providers (n=17) including physicians, advance practice providers, nurses, and emergency medical services (EMS) providers. Research data were obtained through chart review (n= 39), survey (n= 31), registry review (n= 27), and interviews (n= 14), with some studies using more than one data collection modality.

In studies focused on patients (vs health care providers) who use POLST (n= 74), race or ethnicity was analyzed in 45 (58.4%), and 15 (33.3%) had statistically significant differences in POLST outcomes between racial or ethnic groups (See Supplementary Table S3). Among those studies that measured race or ethnicity among POLST users (n= 43), 15 (45%) reported a statistically significant difference in POLST use between racial or ethnic groups. Overall, study participants who were white or Caucasian were more likely to have a POLST form than other racial or ethnic groups.^18–23 Additionally, white and non-Hispanic individuals were more likely to have less aggressive treatment orders (e.g., DNR, comfort measures only, or limited treatment) when compared to other racial and ethnic groups.^18,23–27

3.2. Quality Appraisal

No randomized controlled trials or systematic reviews with meta-analysis were retrieved through our search strategy. Therefore, none of the included studies were appraised as the highest rated study design and resulting quality of evidence (level 1). The highest quality of evidence identified were prospective cohort studies (level 2, n=2). Most included studies were case-control or retrospective cohort study (level 3; n= 38) or case series, cross-sectional, or qualitative study (level 4; n=54). Opinion pieces and case studies (level 5) were not retrieved based on inclusion and exclusion criteria. Quality of evidence is reported for each manuscript in the Supplemental Materials (Tables S1 and S2).

3.3. Outcomes of POLST

Fifteen studies examined use of POLST in both a group of patients who used POLST (i.e., the “intervention” arm) and a group of patients who did not (i.e., the “control” arm) and were observational studies following a cohort or case-control design (level 3 quality of evidence). However, one study did not do a between-group analysis between users and non-users of POLST (i.e., described the sample in terms of descriptive statistics only); therefore outcomes from that study were not counted.²² Overall, there were 70 total outcomes assessed across the 15 studies. When organized based on the ACP Outcomes Framework,¹⁰ there were no studies with Process outcomes. There were 12 with Action, 19 with Quality of Care, 4 with Health Status, and 35 with Healthcare Utilization outcomes. Overall, 40/70 (57%) of outcomes were statistically significantly different (p<0.05) between POLST users and non-POLST users. Table 1 shows all measured outcomes by outcome domain and subdomain.

Table 1.

Association of POLST use with Outcomes by ACP Domain.

Outcome Domain	Subdomain	Outcome Details	Outcomes No.	Sig. Associations No. (%)
TOTAL			70	40 (57)
Process	Overall		0	0
Action	Overall		12	9 (75)
	Communication	Recall of conversation about life-sustaining treatments33	1	0 (0)
	Documentation of treatment preferences	DNR or DNI written,34,35 time to DNR/DNI,34 Any Life- Sustaining Treatment Orders,18 Any non-CPR Life- Sustaining Treatment Orders,18 CPR Status Orders,18 Medical Interventions Orders,18 Antibiotic Use Orders,18 Feeding Tube Use Orders,18 DNR Orders,18 Full Code Status Orders,35 Documentation of discussion and informed refusal of intubation36	11	9 (82)
Quality of Care	Overall		19	15 (79)
	Treatment concordant with documentation	Length of stay,18,29 Location of death,28,29,31,32 hospice use,18,32 any aggressive or life-sustaining medical treatment,18 Recitation ceased or not attempted,30 Transported to ED,30 Hospital admission,30 Advanced airway placement,30 Intravenous placement,30 Intraosseous placement,30 Epinephrine administration,30 Cardioversion/Defibrillation,30 Discharge Disposition30	18	14 (78)
	Preferences concordant with documentation*	Identified via medical record review and resident or surrogate review33	1	1 (100)
Health Status	Overall		4	0 (0)
	Quality of life	Received any pain medication,18 Received non-opioid pain medication,18 Mean daily morphine equivalent,18 Any treatments for shortness of breath18	4	0 (0)
Healthcare Utilization	Overall		35	16 (46)
	Hospitalization utilization	Hospital admission,32,41 length of stay,19,32,35,41 30-day readmission19	7	3 (43)
	Use of life sustaining treatment	Aggressive or life-sustaining medical treatment,32,41 Urgent vs non-urgent surgery19	3	1 (33)
	Hospice utilization	Hospice enrollment,18,28,38,41 length of hospice enrollment38	5	3 (60)
	ICU utilization	ICU admission,32,34,41 ICU length of stay,35,41 Cost of ICU stay35	6	1 (17)
	Place of death*	Place of death32,37,38,40	4	3 (75)
	Overall health care  expenditures	Health Care Spending for Index Hospitalization,32 Health Care Spending for 1-year after Index Hospitalization,32 Health Care Spending for Index Hospitalization (Subgroup: Most Severely Ill Patients),32 Health Care Spending for Index Hospitalization (Subgroup: Patients alive 1-year after event),32 Total Direct Cost of hospitalization35	5	3 (60)
	Long-term care utilization	Nursing home length of stay18	1	0 (0)
	Palliative care utilization	Palliative Care Consult,38 Time to Palliative Care Consult38	2	1 (50)
	Discharge disposition*	Discharge disposition19	1	1 (100)
	Home health utilization*	Home Health Use32	1	0 (0)

Key: * indicate new subdomains, not specified by Sudore et al.¹⁰or McMahan et al.¹¹

The highest incidence of statistically significant different outcomes between those who used and did not use POLST was for the Quality of Care outcome domain (n= 15; 79%). Treatment was significantly associated with the specific POLST orders (i.e., comfort measures only) as they were documented (n= 14; 78%),^18,28–32 and residents’ or surrogates’ actual preferences were more likely to be concordant with documented preferences among those with POLST orders (n= 1) than those who did not use POLST.³³

The Action outcome domain had the second highest positive rate of significant outcomes between those who did and did not use POLST (n= 9; 75%). Specifically, POLST use was associated with documentation of treatment preferences (n= 9; 82%).^18,34–36 Only one study examined recall of conversations about life-sustaining treatments as a communication outcome and the findings were not significant.³³

Healthcare Utilization was the most frequently measured outcome domain. In studies comparing utilization in patients with and without POLST, 16 of 35 (46%) healthcare utilization outcomes were statistically significant. Specifically, differences were significant in a majority of studies that examined subdomain outcomes including: discharge disposition (n= 1; 100%);¹⁹ place of death (n= 3; 75%); hospice utilization (n= 3; 60%);^18,28,38,41 and health care expenditures (n= 3, 60%).^32,35 Findings were more mixed for the subdomain outcomes of palliative care utilization (n= 1, 50%),³⁸ hospital utilization (n=3, 43%),^19,32,35,41 the use of aggressive or life sustaining treatment (n= 1, 33%);^19,32,41 and ICU utilization (n= 1, 17%).^32,34,35,41

There were no statistically significant associations among 4 Health Status outcomes related to symptom management (e.g., administration of opiates for pain management).¹⁸

4. Discussion

Our findings suggest that overall, POLST use is significantly associated with key ACP outcomes including Quality of Care (e.g., goal concordant care), Action outcomes (e.g., discussions and documentation), and Healthcare Utilization (e.g., place of death and hospice use). The Quality of Care domain had the highest number of significant outcomes including treatment concordance with documentation, and preferences concordant with documentation. However, several other measures in this domain have not been included in studies comparing POLST using and non-POLST using populations, such as satisfaction with communication, decision-making, or medical care as well as congruence between patient + surrogate/patient + health care provider.^10,11 Action outcomes measuring completion of specific ACP tasks were also largely statistically significant, such as documentation of treatment preferences. This ACP outcome is intended to address patient related ACP tasks, which is an imperfect fit with POLST because although it is a tool that can be used to document the outcomes of ACP discussions, it is also a medical order that requires clinician involvement. Most states require a patient or surrogate signature on POLST, it is possible that at least in some situations, the patient did not recall the conversation and was unaware there was a POLST on file for a variety of reasons.^33,42,43 Further research is needed in the Action domain to assess communication with surrogates, ideally building on existing qualitative work.^42,44,46

The largest number of outcomes measured fell into the Healthcare Utilization domain and about half were significant, suggesting POLST is associated with measures including hospice use, place of death, and the use life-sustaining treatment. It is notable that outcomes in this category were extracted from studies that did not specify the orders on POLST and instead looked generally at POLST use, grouping together patients with POLST irrespective of the specific orders on POLST. This likely diluted the effect that would be seen if the focus was on whether treatments were concordant with specific POLST orders, as it groups patients with orders for full treatment together with patients who have orders to limit treatment. This focus on utilization as an ACP outcome is similar to the findings of a recent scoping review of the pragmatic characteristics of ACP outcomes in dementia clinical trials, which found that nearly half (49%) of outcomes across all included studies measured healthcare utilization.⁴⁷ This focus may reflect an expectation that concordance between healthcare utilization and ACP is an adequate proxy for goal concordant care. However, this assumption is problematic without verifying that documentation reflects preferences at the time care is indicated.^48–50 Recent studies on POLST preference concordance (captured in Quality of Care) suggest that while patients with POLST are more likely than patients without POLST to have documentation about code status and hospitalizations that reflects current medical preferences,³³ discordance between current preferences and existing POLST orders still exists for a variety of reasons.^{43, 70} Moreover, some of the utilization measures are inconsistent with the goals of POLST (e.g., health care expenditures) or of unclear value in evaluating POLST (e.g., palliative care utilization). There was only one study focused on Health Status (e.g., quality of life, mental health) and were no studies focused on Process outcomes such as readiness, self-efficacy, and knowledge, though knowledge has been examined in studies without comparison groups of non-POLST users.^43,51,52

This review highlights both strengths of the field and areas for future research. Despite a growing body of research on POLST, the overall level of evidence (or certainty of findings) as assessed by study design is modest and our findings about the quality of the evidence following this broader assessment of ACP outcomes are consistent with the conclusions of prior review focused on treatment limitations.⁹ There are no randomized controlled trials and only a minority included comparison groups of non-POLST using participants, limiting the number of studies that could be included in this review. Outcomes from studies that only focused on patients with POLST were excluded because it is difficult to draw conclusions about the effect of POLST in the absence of comparison groups, though many of these studies contain valuable information about the performance of POLST. Among the studies with comparison groups included in this outcomes assessment, the use of POLST is uncontrolled and in at least some studies, potential group differences were not adjusted for within these comparisons. This may reflect, in part, challenges in restricting POLST use in the research context or conversely requiring use, the difficulty with randomizing given that POLST may be introduced to an eligible patient by many different health care providers, and broader system factors that impact whether POLST is honored. Findings reflect associations and do not permit causal inferences. There is a need for more rigorous study design including prospective mixed methods studies as well as high-quality pragmatic trials to identify best practices and improve POLST quality.

Future research should address the full breadth of potential ACP outcomes as they relate to POLST use. Specifically, more research is needed on issues related to Action and Quality of Care and strategies to support high quality communication and decision-making. This includes a focus on the extent to which POLST orders are concordant with both current patient preferences and treatments received, which would provide a true assessment of concordance only achieved by one study in this review.³⁰ Importantly, statistically significant outcomes identified in this review may still represent areas where there is room for improvement, and there is a clear need for methodologically robust studies on POLST.

This review has several limitations. First, by limiting our literature search to PubMed and CINAHL databases, we may have missed manuscripts in journals that are not indexed in these databases. We may have also missed manuscripts which exclusively used the acronym of a state POLST program (e.g., TPOPP, MOLST, etc.) but did not spell out the full terms. Second, our evaluation of study quality is limited to level of evidence based on study design and does not take a nuanced look at each study; this was necessary due to the tremendous heterogeneity of POLST research identified in our review. Thirdly, our analysis focused on the available evidence, which consists of observational studies. Since there were no RCTs, we choose to include only studies with a comparison group of non-POLST users. Studies focused on POLST using patients only are of clear value but lack context to assess whether POLST is better or worse than usual care. Finally, it is possible that study outcomes are affected in part by differences in resources available to support implementation and on-going education within the study state. Assessing state level differences in resources was outside the scope of this review and infeasible given the range of dates included in this review.

5. Conclusion

Findings of this review indicate that POLST is associated with patient outcomes including Quality of Care (e.g., goal-concordant care) and Action outcomes (e.g., communication and documentation), though the available level of evidence is modest. Future research on POLST should center on high-quality designs with well-characterized study populations, intervention fidelity, and implementation strategies to reach representative populations and measure a broad range of that assess a broad range of person and health system level outcomes.