Association of Physician Orders for Life-Sustaining Treatment With ICU Admission Among Patients Hospitalized Near the End of Life

Key Points

Question  For patients with treatment-limiting Physician Orders for Life-Sustaining Treatment (POLST) hospitalized near the end of life, how often is their inpatient care consistent with POLST-ordered limitations?

Findings  In this retrospective cohort study of 1818 decedents with POLSTs who were hospitalized within 6 months of death, rates of intensive care unit (ICU) admission differed significantly by POLST order for medical interventions (31% for those who indicated “comfort measures only,” 46% for those who indicated “limited additional interventions,” and 62% for those who indicated “full treatment”).

Meaning  For patients hospitalized near the end of life, treatment-limiting POLSTs were associated with significantly lower rates of ICU admission compared with full-treatment POLSTs, although many patients with treatment-limiting POLSTs received care that was potentially discordant with their POLST.

Importance  Patients with chronic illness frequently use Physician Orders for Life-Sustaining Treatment (POLST) to document treatment limitations.

Objectives  To evaluate the association between POLST order for medical interventions and intensive care unit (ICU) admission for patients hospitalized near the end of life.

Design, Setting, and Participants  Retrospective cohort study of patients with POLSTs and with chronic illness who died between January 1, 2010, and December 31, 2017, and were hospitalized 6 months or less before death in a 2-hospital academic health care system.

Exposures  POLST order for medical interventions (“comfort measures only” vs “limited additional interventions” vs “full treatment”), age, race/ethnicity, education, days from POLST completion to admission, histories of cancer or dementia, and admission for traumatic injury.

Main Outcomes and Measures  The primary outcome was the association between POLST order and ICU admission during the last hospitalization of life; the secondary outcome was receipt of a composite of 4 life-sustaining treatments: mechanical ventilation, vasopressors, dialysis, and cardiopulmonary resuscitation. For evaluating factors associated with POLST-discordant care, the outcome was ICU admission contrary to POLST order for medical interventions during the last hospitalization of life.

Results  Among 1818 decedents (mean age, 70.8 [SD, 14.7] years; 41% women), 401 (22%) had POLST orders for comfort measures only, 761 (42%) had orders for limited additional interventions, and 656 (36%) had orders for full treatment. ICU admissions occurred in 31% (95% CI, 26%-35%) of patients with comfort-only orders, 46% (95% CI, 42%-49%) with limited-interventions orders, and 62% (95% CI, 58%-66%) with full-treatment orders. One or more life-sustaining treatments were delivered to 14% (95% CI, 11%-17%) of patients with comfort-only orders and to 20% (95% CI, 17%-23%) of patients with limited-interventions orders. Compared with patients with full-treatment POLSTs, those with comfort-only and limited-interventions orders were significantly less likely to receive ICU admission (comfort only: 123/401 [31%] vs 406/656 [62%], aRR, 0.53 [95% CI, 0.45-0.62]; limited interventions: 349/761 [46%] vs 406/656 [62%], aRR, 0.79 [95% CI, 0.71-0.87]). Across patients with comfort-only and limited-interventions POLSTs, 38% (95% CI, 35%-40%) received POLST-discordant care. Patients with cancer were significantly less likely to receive POLST-discordant care than those without cancer (comfort only: 41/181 [23%] vs 80/220 [36%], aRR, 0.60 [95% CI, 0.43-0.85]; limited interventions: 100/321 [31%] vs 215/440 [49%], aRR, 0.63 [95% CI, 0.51-0.78]). Patients with dementia and comfort-only orders were significantly less likely to receive POLST-discordant care than those without dementia (23/111 [21%] vs 98/290 [34%], aRR, 0.44 [95% CI, 0.29-0.67]). Patients admitted for traumatic injury were significantly more likely to receive POLST-discordant care (comfort only: 29/64 [45%] vs 92/337 [27%], aRR, 1.52 [95% CI, 1.08-2.14]; limited interventions: 51/91 [56%] vs 264/670 [39%], aRR, 1.36 [95% CI, 1.09-1.68]). In patients with limited-interventions orders, older age was significantly associated with less POLST-discordant care (aRR, 0.93 per 10 years [95% CI, 0.88-1.00]).

Conclusions and Relevance  Among patients with POLSTs and with chronic life-limiting illness who were hospitalized within 6 months of death, treatment-limiting POLSTs were significantly associated with lower rates of ICU admission compared with full-treatment POLSTs. However, 38% of patients with treatment-limiting POLSTs received intensive care that was potentially discordant with their POLST.

Approximately 20% to 30% of people in the United States are admitted to an intensive care unit (ICU) near the end of life.^1,2 However, many patients with chronic illness prioritize symptom relief over life extension,^3,4 and ICU admission may be potentially unwanted and inconsistent with their goals. To reduce goal-discordant care, one widely implemented approach to documenting treatment preferences in advance is the Physician Orders for Life-Sustaining Treatment (POLST), a portable medical order with treatment limitations for emergency care.⁵ Treatment-limiting POLSTs have been associated with less in-hospital death, hospitalization, and unwanted cardiopulmonary resuscitation (CPR) and with a low incidence of intensive care in nursing home residents.⁶ However, the associations between POLST order for medical interventions and measures of intensive care in other populations are less studied.

This study evaluated the associations between POLST order for medical interventions and ICU admission, as well as the incidence and predictors associated with POLST-discordant intensive care among adults with chronic life-limiting illness hospitalized near the end of life. POLST-discordant care was defined as admission to the ICU when a preexisting POLST form stated that intensive care should be avoided. First, the study examined whether there were differences in intensive care received near the end of life by type of POLST order, hypothesizing that patients with treatment-limiting POLSTs would receive less intensive care. Second, the study examined potential risk factors for POLST-discordant care among patients with treatment-limiting POLSTs.

The study hypothesized that patients with older age, white and non-Hispanic race/ethnicity, higher educational attainment, or who had signed their own POLST would be at lower risk of POLST-discordant care. Additional hypotheses were that patients with more predictable illness trajectories (eg, cancer or dementia) would be less likely to receive POLST-discordant intensive care than patients without these conditions,^7,8 whereas patients with less predictable illnesses (eg, traumatic injury) would be more likely to receive POLST-discordant intensive care.⁹

The study design and a waiver of informed consent for participants were approved by the University of Washington institutional review board.

We conducted a retrospective cohort study of decedents with preexisting POLSTs who were hospitalized near the end of life at the 2 teaching hospitals of UW Medicine, which include a quaternary-care academic medical center and a regional tertiary care and level I trauma center. These hospitals function as a unified health care system, sharing medical staff, house staff, policies, and an electronic health record (EHR). At both hospitals, the decision to admit a patient to the ICU is made jointly by the referring clinician and the admitting ICU physician. All ICU patients are cared for by intensivists in a closed-ICU model. UW Medicine has a mature palliative care program that includes educational and quality improvement initiatives spanning the outpatient, inpatient, and ICU settings.

Participants included all individuals who died in Washington State between January 1, 2010, and December 31, 2017; had chronic life-limiting illness; were hospitalized in the last 6 months of life at a study hospital; and had completed a POLST before their last inpatient admission. We used Washington State death certificates to identify decedents in the study period and linked death certificates to EHR records using social security numbers. Chronic life-limiting illness was defined by having any of 9 chronic conditions documented by diagnosis code in the EHR within the last 2 years of life: cancers with poor prognosis (primary malignancies with poor prognoses, leukemias, and metastatic disease), chronic lung disease, coronary artery disease, congestive heart failure, peripheral vascular disease, chronic renal failure, severe chronic liver disease, diabetes with end-organ damage, and dementia. These conditions are used by the Dartmouth Atlas to study end-of-life care in the United States and are associated with 90% of deaths among Medicare beneficiaries.¹⁰ We excluded patients younger than 18 years at death and those hospitalized for elective surgery.

The Washington State POLST program is a mature, widely disseminated intervention endorsed by the National POLST Paradigm since 2005.^11,12 The Washington State POLST is scanned as a paper document into the EHR and is recognized as a medical order in all care settings, including hospitals. While it is customary for hospital personnel to replace POLSTs with hospital-specific resuscitation orders on admission, in-hospital treatment is often guided by POLST orders. POLSTs provide guidance and legal protection to health care workers who follow them, but they are not inviolable; patients or their surrogates may void the POLST and request alternative treatment at any time. Although a POLST may be completed for any patient, POLSTs are intended for patients with “serious chronic or acute illness nearing its end stages or likely to progress to a life-threatening state suddenly.”¹² All states’ implementation of POLSTs have an order for whether or not to perform CPR in the event of cardiopulmonary arrest and an order for medical interventions for patients who are ill but not in cardiopulmonary arrest.¹¹ In Washington State, the options for medical interventions orders are “full treatment,” “limited additional interventions” (hereafter, limited interventions), or “comfort measures only” (hereafter, comfort only) (eFigure 1 in the Supplement).

For this study, POLST order for medical interventions, signatory, and completion date were manually abstracted from scanned POLST forms. Abstractors were blinded to study outcomes, and the last POLST preceding the patient’s last hospitalization was used for analysis. Because POLSTs are not used to document real-time changes in inpatients’ treatment limitations, we did not analyze POLSTs completed after hospital admission.

All outcomes were measured during the last hospitalization of life and treated as binary variables. Intensive care was defined in 2 ways: (1) admission to the ICU, and (2) receipt of any of 4 prespecified life-sustaining treatments: mechanical ventilation, vasoactive infusions (dobutamine, dopamine, epinephrine, isoproterenol, milrinone, norepinephrine, phenylephrine, or vasopressin), new dialysis or continuous renal replacement therapy, or CPR. These outcomes were electronically abstracted from structured clinical documentation in the EHR. We considered receipt of CPR anywhere in the hospital to be equivalent to ICU admission. Acute care patients boarding in the ICU because of bed capacity were not considered to have been admitted to the ICU.

Among all patients with POLSTs, we evaluated the association of POLST order for medical interventions with the primary outcome of ICU admission and the secondary outcome of receipt of 1 or more aforementioned life-sustaining treatments.

Among patients with treatment-limiting POLST order for medical interventions, we evaluated risk factors for the outcome of POLST-discordant intensive care. For patients with comfort-only POLSTs, POLST-discordant intensive care was defined as any ICU admission except admissions for symptom management only. For patients with limited-interventions POLSTs, POLST-discordant intensive care was defined as any ICU admission except admissions for symptom management only and admissions solely for delivery of noninvasive ventilation without additional life-sustaining treatments as specified by the Washington State POLST (eFigure 1 in the Supplement). To identify ICU admissions for symptom management only, we manually reviewed clinician documentation of treatment plans for patients in whom a standardized comfort-care order set was initiated within 24 hours of ICU admission.

Age, race/ethnicity, educational attainment, and date and location of death were obtained from death certificates. Death certificate race and ethnicity fields are categorical and typically collected from a family member or other informant; this covariate was included because of previously reported differences in end-of-life care by race/ethnicity. Educational attainment was converted from an 8-category variable to approximate years of education. Chronic life-limiting illnesses, including histories of cancers with poor prognosis and dementia, were measured using diagnosis codes documented over decedents’ last 2 years of life.¹⁰ For patients with treatment-limiting POLSTs, the reason for hospital admission was determined by manual chart abstraction; we created a binary variable of admission for traumatic injury vs all other admitting diagnoses. Date of POLST completion was manually abstracted from the POLST as the date accompanying the patient or surrogate’s signature; if this date was not completed, we used the earliest written date on the POLST (typically the date of the clinician’s signature). POLSTs without legible dates were assumed to have been signed on the date they were scanned into the EHR. POLST signatory (patient vs surrogate) was manually abstracted from the signer’s printed name or signature; illegible names/signatures were treated as missing data.

To evaluate the association between the primary outcome of ICU admission and POLST order for medical interventions (comfort only, limited interventions, full treatment), we used modified Poisson regression¹³ without and with adjustment for age at admission, race/ethnicity, years of education, log-transformed days from POLST completion to admission, history of cancer with poor prognosis, history of dementia, and POLST signatory. POLST order was modeled as a nominal exposure, with full treatment as the referent category. Age, years of education, and log-transformed days from POLST completion to admission were modeled continuously. All other exposures were modeled as binary variables. Similar analyses were performed for the secondary outcome of receipt of life-sustaining treatments. We explored associations between POLST order and individual life-sustaining treatments (eg, mechanical ventilation) using the same approach.

To evaluate potential risk factors for POLST-discordant care among patients with treatment-limiting POLSTs, we used modified Poisson regression¹³ without and with adjustment for age at admission, race/ethnicity, years of education, log-transformed days from POLST completion to admission, history of cancer with poor prognosis, history of dementia, POLST signatory, and admitting diagnosis of traumatic injury. Age, years of education, and log-transformed days from POLST completion to admission were modeled continuously, and all other exposures were modeled as binary variables. We performed separate analyses by POLST order (comfort only vs limited interventions). Additionally, to explore temporal trends in POLST-discordant care, we performed a secondary analysis that included date of death as a continuous variable using a similar approach.

We used modified Poisson regression with robust (sandwich) estimation of variance applied to binary, nonrepeated outcomes¹³ instead of logistic regression to allow estimation of relative risk when the rare disease assumption is violated. We used robust (sandwich) estimation of variance, since the mean-variance relationship of the Poisson distribution may be inappropriate for our binary outcomes. Multiple imputation with chained equations was used to impute missing data for race/ethnicity, educational attainment, and POLST signatory. We assumed that data were missing at random and modeled missing data using logistic or linear regression depending on the covariate, as conditioned on the remaining exposures and outcome for each model, stratified by POLST order when appropriate. We ran 20 imputations a priori and assessed adequacy of number of imputations using the largest fraction of missing information for each model. Effect estimates were combined using Rubin rules. Parameter estimates were exponentiated to obtain estimates of relative risk.

Estimates of incidence and relative risks refer to the period from study hospitalization to death, assuming intensive care is not received at another hospital after discharge from the study hospital. To address the sensitivity of results to the latter assumption, we performed additional analyses restricted to patients who died during the study hospitalization.

P < .05 (2-sided) was considered statistically significant. Because of the potential for type I error due to multiple comparisons, findings from analyses of secondary end points should be interpreted as exploratory. Analyses were performed using Stata version 16.0 (StataCorp LLC).

Between January 1, 2010, and December 31, 2017, there were 30 803 potentially eligible adult decedents in Washington State with chronic life-limiting illness who received care through UW Medicine, of whom 14 370 were hospitalized in the last 6 months of life at a study hospital. Of these, 1818 (12.7%) had a POLST that preceded the study hospitalization (Figure 1).

Among 1818 decedents with POLSTs hospitalized near the end of life, the mean age was 70.8 (SD, 14.7) years; 752 (41%) were women; and 450 (26%) were nonwhite or Hispanic (Table 1). Most decedents had more than 1 chronic life-limiting illness (median, 3 [interquartile range, 2-4); 39% had cancer, and 26% had dementia. Of 1818 decedents, 656 (36%) had POLST orders for full treatment, 761 (42%) had orders for limited additional interventions, and 401 (22%) had orders for comfort measures only. Missing data were present for race/ethnicity (5%), education (9%), and POLST signatory (14%) (eTable 1 in the Supplement).

POLSTs were completed a median of 111.5 (interquartile range, 28-369) days before hospital admission. At least 1096 POLSTs (60%) were signed by the patient, and at least 461 (25%) were signed by surrogates; 261 (14%) had illegible signatures. Thirty-two POLSTs (1.8%) bore no legible signature date, although all preceded the study hospitalization based on the date they were scanned into the EHR. Among patients with treatment-limiting POLST orders, 13% were admitted to the hospital for traumatic injury. Other reasons for hospital admission are summarized in eTable 2 in the Supplement.

Among decedents with any POLST order for medical interventions, 878 (48%) were admitted to the ICU during the study hospitalization. The incidence of ICU admission by POLST order was 123/401 (31%) for those with comfort-only orders, 349/761 (46%) for those with limited-interventions orders, and 406/656 (62%) for those with full-treatment orders (Table 2).

Compared with patients with full-treatment POLSTs, patients with comfort-only or limited-interventions POLSTs were significantly less likely to be admitted to the ICU (comfort only: unadjusted risk, 123/401 [31%] vs 406/656 [62%], adjusted relative risk [aRR], 0.53 [95% CI, 0.45-0.62; P < .001]; limited interventions: unadjusted risk, 349/761 [46%] vs 406/656 [62%], aRR, 0.79 [95% CI, 0.71-0.87; P < .001]) and were also significantly less likely to receive life-sustaining treatments (comfort only: unadjusted risk, 55/401 [14%] vs 279/656 [43%], aRR, 0.38 [95% CI, 0.29-0.49; P < .001]; limited interventions: unadjusted risk, 149/761 [20%] vs 279/656 [43%], aRR, 0.53 [95% CI, 0.44-0.63; P < .001]) (Figure 2). In secondary analyses examining individual life-sustaining treatments, treatment-limiting POLST orders were significantly associated with lower likelihoods of mechanical ventilation, vasoactive infusions, or CPR (eFigure 2 in the Supplement).

Among decedents with treatment-limiting POLST orders for medical interventions, ICU admission occurred in 123 of 401 (31%) with comfort-only orders and 349 of 761 (46%) with limited-interventions orders (Table 2). Five (1%) of these ICU admissions were for the sole purpose of symptom management. Although the Washington State POLST allows delivery of noninvasive ventilation to patients with POLST order for limited interventions, only 31 of 761 patients (4%) with limited-interventions POLSTs were admitted to the ICU for noninvasive ventilation without receiving other life-sustaining treatments.

The incidence of POLST-discordant intensive care was 30% (95% CI, 26%-35%) in the comfort-only group and 41% (95% CI, 38%-45%) in the limited-interventions group, for a combined incidence of 38% (95% CI, 35%-40%). The incidence of POLST-discordant delivery of life-sustaining treatments was 14% (95% CI, 11%-17%) in the comfort-only group and 20% (95% CI, 17%-23%) in the limited-interventions group, for a combined incidence of 18% (95% CI, 15%-20%). There was no evidence that the incidence of POLST-discordant intensive care decreased significantly over the 8 years of the study (comfort only: aRR, 1.01 per year [95% CI, 0.94-1.09; P = .70]; limited interventions: aRR, 1.00 per year [95% CI, 0.96-1.04; P = .90]) (eFigure 3 in the Supplement).

In evaluating potential risk factors for POLST-discordant intensive care (Figure 3), patients with a history of cancer were significantly less likely to receive POLST-discordant care compared with patients without cancer in both the comfort-only and the limited-interventions groups (comfort only: unadjusted risk, 41/181 [23%] vs 80/220 [36%], aRR, 0.60 [95% CI, 0.43-0.85; P = .004]; limited interventions: unadjusted risk, 100/321 [31%] vs 215/440 [49%], aRR, 0.63 [95% CI, 0.51-0.78; P < .001]). Additionally, among patients with comfort-only POLSTs, those with dementia were significantly less likely to receive POLST-discordant care compared with patients without dementia (unadjusted risk, 23/111 [21%] vs 98/290 [34%]; aRR, 0.44 [95% CI, 0.29-0.67; P < .001]); this association was not significant in those with limited-interventions POLSTs (unadjusted risk, 99/213 [46%] vs 216/548 [39%]; aRR, 1.02 [95% CI, 0.84-1.25; P = .82]). In both the comfort-only and the limited-interventions groups, patients admitted for traumatic injury were significantly more likely to receive POLST-discordant care compared with patients without traumatic injury (comfort only: unadjusted risk, 29/64 [45%] vs 92/337 [27%], aRR, 1.52 [95% CI, 1.08-2.14; P = .02]; limited interventions: unadjusted risk, 51/91 [56%] vs 264/670 [39%], aRR, 1.36 [95% CI, 1.09-1.68; P = .005]). Among patients with limited-interventions POLSTs, older age was also associated with significantly less POLST-discordant care (aRR, 0.93 per 10 years [95% CI, 0.88-1.00; P = .04]); this association was not observed in patients with comfort-only POLSTs (aRR, 1.01 per 10 years [95% CI, 0.91-1.13; P = .81]). There were no significant associations between POLST-discordant care and non-white race or Hispanic ethnicity, POLST signatory, or days from POLST completion to admission (Figure 3).

Sensitivity analyses restricted to patients who died during the study hospitalization yielded similar associations with expected changes in precision (eFigures 4 and 5 in the Supplement).

In this study of patients with POLSTs hospitalized near the end of life, rates of ICU admission and life-sustaining treatments differed significantly by POLST order for medical interventions. Those with full-treatment orders were significantly more likely to receive ICU admission and life-sustaining treatments than those with comfort-only or limited-interventions orders. However, among those with treatment-limiting POLST order for medical interventions, 38% were admitted to the ICU and 18% received mechanical ventilation, vasoactive infusions, dialysis, or CPR.

While several previous studies of patients with POLSTs in nursing homes have reported a much lower incidence of POLST-discordant medical treatments,^5,14,15 these studies did not specifically examine care received near the end of life. Studies that specifically examine end-of-life care have had mixed findings. The findings of this study are consistent with those from a prior study of 58 decedents with POLSTs in a comprehensive elder care program, which reported a similar incidence of POLST-discordant care.¹⁶ In contrast, 2 prior decedent studies in which POLSTs had been implemented as part of community-wide advance care planning initiatives reported a very low occurrence of POLST-discordant care near the end of life.^17,18 These differences are likely explained by differences in sample populations, approaches to POLST completion, and periods for assessing discordant care. In contrast to the structured implementation of POLSTs within interdisciplinary community-wide advance care planning initiatives represented in these prior studies, this study enrolled a diverse group of patients at a large referral health system, where most POLSTs are completed in the course of usual care by clinicians from both UW Medicine and the surrounding community.

Although treatment-limiting POLST orders were significantly associated with lower intensity of care compared with full-treatment POLST orders, 38% of patients with treatment-limiting POLSTs received potentially unwanted intensive care near the end of life. The experiences of ICU patients and survivors suggest that patients receiving aggressive life-sustaining treatments experience a high burden of unrelieved physical symptoms and emotional distress,^19,20 a finding echoed in the perspectives of family members of patients who die in the ICU.^21,22 Although studies of ICU survivors report that most patients would elect to receive intensive care again, these studies have not examined patients with treatment-limiting POLST forms.^23,24 Additionally, intensive care and mechanical ventilation account for a disproportionate amount of health care expenditures within the last years of life.^1,25 For patients who do not want aggressive treatments near the end of life, unwanted intensive care may incur physical, emotional, and financial costs while providing little value. Reducing unwanted intensive care near the end of life represents an opportunity to simultaneously improve patient-centered care and reduce costs.²⁶

In evaluating the study outcome, it is important to highlight that differences between POLST-discordant care and goal-discordant or inappropriate care may exist for several reasons. In the setting of acute illness, patients, surrogates, and clinicians may encounter circumstances that ethically compel a different treatment course than that outlined by a previously completed POLST; accordingly, many patients with treatment limitations are willing to grant leeway to future decision-makers.²⁷ Nevertheless, it is possible that many cases of POLST-discordant care represent potentially inappropriate care near the end of life, and further study is needed to understand why this happens.

Four factors were identified as independently associated with POLST-discordant intensive care. First, dementia was associated with significantly lower risk of POLST-discordant care in patients with comfort-only POLSTs. This finding is consistent with prior studies showing increased effectiveness of advance directives in reducing intensity of end-of-life care among patients with cognitive impairments.²⁸

Second, cancer was associated with significantly lower risk of POLST-discordant intensive care. This is also consistent with prior studies.²⁹ Although patients with cancer complete POLST forms later in their disease course than those with organ failure or dementia,³⁰ the association observed was independent of POLST timing.

Third, patients admitted for traumatic injury were significantly more likely to receive POLST-discordant care. This finding complements a prior study that found no difference in trauma-related ICU admission rates between patients with and without treatment-limiting POLSTs.³¹ These findings may be related to the time pressure of trauma care; additionally, traumatic injury may be an uncommon scenario considered in the shared decision-making that informs a patient’s POLST order.³² Although POLSTs are designed to apply to all clinical circumstances, divergence of actual acute illness (eg, hip fracture) from the anticipated clinical scenarios that inform POLST order (eg, exacerbation of chronic disease) may lead patients, surrogates, or clinicians to choose more aggressive care than that designated by POLST order.^33,34

Fourth, in adjusted analyses, older age was independently associated with a significant decrease in POLST-discordant care among those with limited-interventions POLSTs but not among those with comfort-only POLSTs. This finding is consistent with prior studies that have found older age to be associated with decreased intensity of care.^35,36

The results of this study support the overall hypothesis that patients with less predictable prognostic trajectories^7,8 may receive more intense and more POLST-discordant care near the end of life. This may be because of prognostic uncertainty biasing clinicians and decision-makers toward more aggressive care, unanticipated clinical developments that bring the applicability of POLST orders into question, prevailing care processes for specific illnesses that favor aggressive care, or differences in care that facilitate or impede the communication and interpretation of patients’ values, goals, and treatment preferences.^33,34,37

There were no significant associations between POLST-discordant care and race/ethnicity, educational attainment, or decision-making capacity at the time of POLST completion (assessed by whether the patient was able to sign his or her own POLST). Minority race and ethnicity have been associated with less advance directive completion and greater intensity of end-of-life care,³⁸ although there is emerging evidence that these differences may be explained by differences in care delivery.³⁹ The racial and ethnic composition of this study sample limits the ability to draw conclusions that generalize to other settings. There was also no significant association between POLST-discordant intensive care and date of death over the 8-year study period, suggesting that increasing experience with POLST forms is not reducing POLST-discordant intensive care.

This study has several limitations. First, by enrolling patients with POLSTs hospitalized near the end of life and measuring POLST-discordant care in the hospital, these findings do not apply to patients with POLSTs who are not hospitalized near the end of life (as may be the case when care is aligned with a comfort-only POLST). Therefore, the results of this study may overestimate the incidence of POLST-discordant care among all patients with POLSTs.

Second, the study only included POLST-discordant care that occurred at the study hospitals and may misclassify patients who received POLST-discordant care at another hospital or from emergency medical services without subsequent transfer to a study hospital. Such misclassification would result in underestimation of the incidence of POLST-discordant care from the time of study hospitalization to death.

Third, by enrolling decedents, the study is unable to evaluate POLST-discordant care delivered to individuals who go on to survive their critical illness. This concern is tempered by the high mortality of critical illness among individuals with chronic life-limiting illness; additionally, examining decedents allowed the study to specifically evaluate care received near the end of life—a critical period for the intended function of the POLST.⁵

Fourth, the study only included POLSTs on file at the study hospitals. However, the observed uptake of POLSTs in this study (12.7%) is similar to statewide estimates of POLST uptake (13%) (Jessica Martinson, MS, Washington State Medical Association, email communication, January 2019).

Fifth, data from death certificates, claims data, and the EHR are susceptible to misclassification.^40,41 To mitigate this, all outcome data were corroborated using multiple sources (eg, by linking claims and clinical data or by corroborating with manual abstraction).

Sixth, analyses did not account for differences in care between the 2 hospitals within UW Medicine. Because each study hospital houses unique subspecialty centers (eg, cancer center, trauma center), hospital site is in the causal pathway for several exposures of interest. Because both hospitals share a common medical staff, policies, management, and culture, the observed differences in care are more likely to be due to patient differences.

Among patients with POLSTs and with chronic life-limiting illness who were hospitalized within 6 months of death, treatment-limiting POLSTs were significantly associated with lower rates of ICU admission compared with full-treatment POLSTs. However, 38% of patients with treatment-limiting POLSTs received intensive care that was discordant with their POLST.

Corresponding Author: Robert Y. Lee, MD, MS, Division of Pulmonary, Critical Care, and Sleep Medicine, Cambia Palliative Care Center of Excellence, Harborview Medical Center, University of Washington, 325 9th Ave, Campus Box 359762, Seattle, WA 98104 (rlee06@uw.edu).

Accepted for Publication: January 20, 2020.

Published Online: February 16, 2020. doi:10.1001/jama.2019.22523

Author Contributions: Drs Lee and Kross had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Lee, Brumback, Sathitratanacheewin, Lober, Engelberg, Curtis, Kross.

Acquisition, analysis, or interpretation of data: Lee, Brumback, Sathitratanacheewin, Modes, Lynch, Ambrose, Sibley, Vranas, Sullivan, Engelberg, Curtis, Kross.

Drafting of the manuscript: Lee, Sathitratanacheewin, Engelberg.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Lee, Sathitratanacheewin, Engelberg.

Obtained funding: Lee, Sathitratanacheewin, Curtis.

Administrative, technical, or material support: Sathitratanacheewin, Lober, Modes, Lynch, Sibley, Engelberg.

Supervision: Brumback, Curtis, Kross.

Conflict of Interest Disclosures: Dr Lee reported receiving grants from the National Institutes of Health (NIH). Dr Brumback reported receiving grants from the NIH. Dr Sathitratanacheewin reported receiving a grant from the Prince Mahidol Youth Program Award. Dr Modes reported receiving grants from the NIH. Dr Lynch reported receiving grants from the NIH. Dr Vranas reported receiving grants from the NIH, the Oregon Health & Science University Medical Research Foundation, and the Collins Medical Trust. Dr Sullivan reported receiving grants from the NIH, the Oregon Health & Science University Medical Research Foundation, the American Lung Association, the American Thoracic Society, the Borchard Foundation, and the Knight Cancer Institute. Dr Engelberg reported receiving grants from the NIH, the Cambia Health Foundation (funding from which supports Drs Brumback and Lober and Mr Sibley), the National Palliative Care Research Center, the Gordon and Betty Moore Foundation, the Stupski Foundation, and the Cystic Fibrosis Foundation. Dr Curtis reported receiving grants from the NIH, the Cambia Health Foundation (funding from which supports Drs Brumback and Lober and Mr Sibley), and the National Palliative Care Research Center. Dr Kross reported receiving grants from the NIH, the American Lung Association, and the American Thoracic Society. No other disclosures were reported.

Funding/Support: This study was supported by the National Institutes of Health, Cambia Health Foundation, and UW Medicine. Dr Lee was supported by an F32 award (HL142211) and a K12 award in implementation science (HL137940); Drs Lee and Modes, by a palliative care T32 training fellowship (HL125195); and Drs Modes and Lynch, by a pulmonary/critical care T32 training fellowship (HL007287), all from the National Heart, Lung, and Blood Institute (NHLBI). Dr Sathitratanacheewin was supported by the Prince Mahidol Youth Program Award, Bangkok, Thailand. Dr Vranas was supported by a K12 award (HL133115) jointly funded by the NHLBI and the National Institute of Mental Health. Infrastructure and chart abstraction support was provided by the University of Washington Institute of Translational Health Sciences (ITHS), which is funded by the National Center for Advancing Translational Sciences through the Clinical and Translational Science Awards (CTSA) Program (UL1 TR002319).

Role of the Funder/Sponsor: The National Institutes of Health, Cambia Health Foundation, and UW Medicine had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Meeting Presentation: Presented at the Society of Critical Care Medicine 49th Critical Care Congress; February 16, 2020; Orlando, Florida.

Additional Contributions: We acknowledge Lauren Bartlett, BS, Michael Donahue, BS, and Barbara Burke, BS (ITHS), as well as Ross Burnside (Cambia Palliative Care Center of Excellence), for their assistance in digitizing POLST orders. The investigators provided financial compensation via a standard contract for research coordination support services to ITHS for Ms Bartlett, Mr Donahue, and Ms Burke’s work. Mr Burnside received no financial compensation for his contributions.