Differences in End-of-Life Care in the ICU Across Patients Cared for by Medicine, Surgery, Neurology, and Neurosurgery Physicians

Transcription

1 CHEST Original Research Differences in En-of-Life Care in the ICU Across Patients Care for y Meicine, Surgery, Neurology, an Neurosurgery Physicians Erin K. Kross, MD ; Ruth A. Engelerg, PhD ; Lois Downey, MA ; Joseph Cuschieri, MD ; Matthew R. Hallman, MD ; W. T. Longstreth Jr, MD, MPH ; Davi L. Tirschwell, MD ; an J. Ranall Curtis, MD, MPH, FCCP CRITICAL CARE Backgroun: Some of the challenges in the elivery of high-quality en-of-life care in the ICU inclue the variaility in the characteristics of patients with certain illnesses an the practice of critical care y ifferent specialties. Methos: We examine whether ICU attening specialty was associate with quality of en-of-life care y using ata from a clustere ranomize trial of 14 hospitals. Patients ie in the ICU or within 30 h of transfer an were categorize y specialty of the attening physician at time of eath (meicine, surgery, neurology, or neurosurgery). Outcomes inclue family ratings of satisfaction, family an nurse ratings of quality of ying, an ocumentation of palliative care in meical recors. Associations were teste using multipreictor regression moels ajuste for hospital site an for patient, family, or nurse characteristics. Results: Of 3,124 patients, the majority were care for y an attening physician specializing in meicine (78%), with fewer from surgery (12%), neurology (3%), an neurosurgery (6%). Family satisfaction i not vary y attening specialty. Patients with neurology or neurosurgery attening physicians ha higher family an nurse ratings of quality of ying than patients of attening physicians specializing in meicine ( P,.05). Patients with surgery attening physicians ha lower nurse ratings of quality of ying than patients with meicine attening physicians ( P,.05). Chart ocumentation of inicators of palliative care iffere y attening specialty. Conclusions: Patients care for y neurology an neurosurgery attening physicians have higher family an nurse ratings of quality of ying than patients care for y meicine attening physicians an have a ifferent pattern of inicators of palliative care. Patients with surgery attening physicians ha fewer ocumente inicators of palliative care. These finings may provie insights into potential ways to improve the quality of ying for all patients. Trial registry: ClinicalTrials.gov; No.: NCT ; URL: CHEST 2014; 145(2): Areviations: DNR 5 o not resuscitate; EOL 5 en of life; QODD 5 Quality of Dying an Death Death is common in the ICU in the Unite States, 1,2 an the importance of integrating quality palliative care into the ICU is eing increasingly recognize. One of the challenges in the elivery of high-quality palliative care in the ICU is the ramatic variaility in en-of-life (EOL) care across ifferent ICUs. 3 Each ICU has its own culture that is shape y many factors, incluing its structure, history, policies, processes of care, an attitues. 4 Other important factors in the culture of the ICU are the types of patients an specialty of proviers in an ICU. The characteristics of patients with certain types of illness or injury an the practice of ifferent types of critical-care physicians can present challenges to the integration of palliative care into the ICU. 5 A few reports have focuse specifically on EOL care of the neurology or neurosurgery patient. 6,7 Devastating neurologic insults often occur suenly in the asence of chronic eilitating conitions, an lifesustaining interventions are often initiate emergently journal.pulications.chestnet.org CHEST / 145 / 2 / FEBRUARY

2 efore a iagnosis or prognosis can e efine. 8 Conversely, patients with many meical an some surgical iagnoses are more likely to have chronic comori illnesses relate to their ICU amission. One prior stuy foun that nurses rate quality of ying higher for neurology an neurosurgery patients than patients of other specialties. 9 However, to our knowlege, a comparison of the quality of EOL care across ifferent physician specialties in the ICU has not otherwise een escrie. To explore ifferences in EOL care across meicine, surgery, neurology, an neurosurgery physicians, we examine a cohort of patients who ie in or shortly after a stay in the ICU. We aske whether the following outcomes iffere y the physician specialty of the attening physician of recor at the time of eath: (1) family or nurse satisfaction with care, (2) family or nurse ratings of quality of ying, an (3) ocumentation of elivery of palliative care. Design Materials an Methos Data were collecte as part of a cluster-ranomize trial esigne to evaluate the efficacy of a multifacete, interisciplinary intervention to improve palliative care in the ICU (the Integrating Palliative an Critical Care stuy). Details of the stuy esign an results of the ranomize trial have een previously reporte All stuy proceures were approve y the institutional review oar at all sites. Stuy Participants All patients who ie in the ICU after a minimum stay of 6 h or within 30 h of transfer from the ICU were eligile for the stuy. Patients with rain eath were exclue. Hospitals in the Seattle- Tacoma, Washington, area were eligile if they ha enough ICU eaths to meet sample size requirements for the Integrating Palliative an Critical Care stuy. 11 Of 16 eligile hospitals, 15 agree to participate. The current stuy inclues the 12 sites from the ranomize trial as well as two of the pilot sites (one site was a pilot for the intervention ut i not inclue chart astraction). These 14 hospitals inclue two university-affiliate teaching hospitals; Manuscript receive June 11, 2013; revision accepte Septemer 16, Affiliations : From the Division of Pulmonary & Critical Care Meicine, Department of Meicine (Drs Kross, Engelerg, an Curtis an Ms Downey), Department of Surgery (Dr Cuschieri), Department of Anesthesiology & Pain Meicine (Dr Hallman), an Department of Neurology (Drs Longstreth an Tirschwell), University of Washington, Harorview Meical Center, Seattle, WA. Funing/ Support: This stuy was supporte y the National Institute of Nursing Research [R01NR05226 to Dr Curtis] an the National Heart, Lung an Bloo Institute [K23HL to Dr Kross]. Corresponence to: Erin K. Kross, MD, Division of Pulmonary & Critical Care Meicine, Harorview Meical Center, 325 Ninth Ave, Box , Seattle, WA 98104; ekross@uw.eu 2014 American College of Chest Physicians. Reprouction of this article is prohiite without written permission from the American College of Chest Physicians. See online for more etails. DOI: /chest three community-ase teaching hospitals; an nine communityase, nonteaching hospitals. Most of the hospitals (12 of 14) ha one ICU (either meicine or mixe meicine-surgical). Of the two remaining hospitals, one ha two ICUs (surgery an neurology) an the other ha six ICUs (trauma, surgical, cariac, meicine, urn, an neurosurgical). The majority of the hospitals (13 of 14) ha a semi-open ICU structure with either optional or require intensivist consultation; the six ICUs at the remaining hospital inclue oth close an open ICU structures. Patients who ie were ientifie using ischarge an transfer logs. Stuy activities were from August 2003 to Feruary Stuy proceures were approve y the institutional review oar at each stuy site (e-appenix 1). Data Collection Family Surveys: Surveys were maile to families of patients who ie uring the stuy perio. Surveys were maile 1 to 2 months after the patient ie an were written in English. One family memer per patient was aske to respon. The survey packet inclue a cover letter, consent form, $10 incentive, postage-pai return envelope, an questionnaire ooklet. The questionnaire ooklet inclue emographic questions, the Quality of Dying an Death (QODD) questionnaire, an the Family Satisfaction in the ICU survey. Survey follow-up use a stanarize approach 13 that inclue reminers sent 2 weeks after the initial mailing an secon survey packets sent after 4 weeks if there was no response to the initial mailing. Nurse Surveys: Nurse questionnaires were istriute within 72 h of eath to the nurse caring for the patient at the time of eath/transfer an the nurse from the prior shift. Survey packets inclue a cover letter, consent form, coffee-car incentive, the QODD questionnaire, an questions asking for ratings of the care the patient receive in the last ays of life. The same proceures were use to follow-up with nonresponents as with family memers. 13 Chart Astraction: Data astractors were traine y two researchastraction trainers. Training inclue a minimum of 80 h of practice astraction followe y reconciliation with trainers. Training continue until 95% agreement was reache with trainers. For ongoing quality control, astracters coreviewe a 5% ranom sample, ensuring at least 95% agreement on the 440 astracte ata elements. Death Certificate Data: Washington State eath certificates were linke y patient ientifier to provie ata that were unavailale or incomplete in the meical recor. Data otaine from eath certificates inclue patient race, eucation, marital status, an cause of eath. Variales of Interest Outcome Measures Quality of Dying an Death Questionnaire Family memers an nurses complete the valiate QODD questionnaire measuring family- or clinician-assesse quality of ying. 9,14-16 For this stuy, we examine a single-item, quality-of-ying rating (range, 0-10) that is associate with ICU palliative care. 17 Higher scores inicate higher-quality ying. Family Satisfaction in the ICU Survey This survey is a valiate 34-item questionnaire measuring family satisfaction with ICU care. 18,19 Scores on 24 items provie a total satisfaction score, as well as two omain scores: satisfaction with care an satisfaction with ecision-making. 20 Scores are recoe an recalirate to a 0 to 100 range, with higher values inicating higher satisfaction Original Research

3 Nurse-Assesse Satisfaction With Care Two questions were use to assess nurse ratings of satisfaction with care of patients an their family. Nurses were aske to rate on a 0 to 10 scale (from worst care possile to est care possile) the care your patient receive in the last several ays of his/her life while in the ICU from all octors an other health-care proviers comine. Nurses were also aske to rate the following on a 0 to 10 scale (from not satisfie at all to very satisfie): How satisfie were you with how well the health-care team met the family s nees while their love one was in the ICU? 21 Chart-Base Inicators of Palliative Care Inicators of palliative care were ientifie from meical recors an inclue aspects of care that have een previously efine in consensus ocuments as inicators of palliative care. 22,23 These inclue palliative care consultation, social work services, spiritual care, o not resuscitate (DNR) orer at time of eath, withholing or withrawal of lifesustaining therapies, pain assessments in the last 24 h of life, avoiance of CPR prior to eath, a family conference within 72 h of amission, a iscussion of prognosis within 72 h of amission, ICU length of stay, an time from ICU amission to withrawal of mechanical ventilation. These inicators of palliative care have een shown to e associate with higher family ratings of quality of ying, 17 higher ratings of family satisfaction with care, 24 an ecrease family psychologic symptoms after eath of their love one, 25 proviing valiation of their usefulness as inicators of quality palliative care. Preictors an Covariates ICU Attening Physician Specialty Patients were categorize y the specialty of the attening physician caring for the patient at the time of eath, efine y the specialty of the attening physician of recor ocumente on the patient s eath summary. We use the following four categories: (1) meicine (family meicine, internal meicine, an internal meicine suspecialties), (2) surgery (general surgery an surgical suspecialties except neurosurgery), (3) neurology, an (4) neurosurgery. Patient, Family, an Nurse Characteristics Patient characteristics were collecte from meical recors an eath certificates. Demographic variales for patients inclue age, sex, race, cause of eath (cancer, trauma, or other), insurance status (insure vs unerinsure), 26 an eucation. Family memer characteristics collecte from family surveys inclue age, sex, race, an relationship to the patient (spouse/partner vs other relationship). Nurse characteristics collecte from nurse surveys inclue age, sex, race, an years of ICU nursing experience. Data Analysis Characteristics of patients, family memers, an nurses were examine y ICU physician specialty using escriptive statistics an expresse as either mean (SD) or numer (percent). Associations etween physician specialty an the outcomes of interest were ase on Toit or roust linear regression moels for family an nurse ratings, Cox regression moels for time-ase variales, an logistic regression moels for ichotomous variales. The choice of Toit or linear regression for pseuocontinuous outcomes (eg, outcomes score 0-10) was ase on the numer of cases at the lowest an highest possile values on the outcome. If 25% of the cases were at either the ceiling or floor, we use Toit regression. To test nurse outcomes, we use clustereregression moels with patients clustere uner nurses. All regression estimates were ase on restricte maximum likelihoo. For Cox moel coefficients, the higher the value, the shorter the associate time perio. A priori, we chose to ajust all moels for hospital (using ummy variales). In aition, any covariate that cause. 20% change in the coefficient for any physician specialty was consiere a con- founer for that preictor-outcome pair an was inclue in that iniviual moel. Covariates that were examine for confouning in all moels inclue patient characteristics (age, sex, race, cause of eath, insurance status, an eucation). For family outcomes, we teste family characteristics (age, sex, race, relationship to patient). For nurse outcomes, we teste nurse characteristics (age, sex, race, an years of ICU nursing experience). Meicine specialty was the reference group in all moels. In the ajuste analyses for each outcome, an overall P value for specialty was calculate ase on the reuction in eviance otaine in a moel in which the coefficients for the three ummy inicators for specialty were freely estimate, when compare with a moel in which the three specialty-relate regression coefficients were constraine to 0.0. Significance was efine as P.05. Results A total of 3,124 patients ie uring the stuy perio at the 14 sites. Of these, 1,185 (38%) ha at least one family-reporte outcome assesse, an 1,198 (38%) ha at least one nurse-reporte outcome assesse. The patients mean age was 69 years, an the majority of patients were non-hispanic white (79%) an male (59%). The majority were care for y a meicine attening physician at the time of eath (78%), with fewer from surgery (12%), neurology (3%), an neurosurgery (6%) ( Tale 1 ). All of the 14 hospitals ha patients with meicine an surgery attening physicians, while eight of the 14 (57%) ha patients with a neurology attening physician an 10 of the 14 (71%) ha patients with a neurosurgery attening physician. A total of 1,184 family memers (mean age, 58 years) respone to the survey. The majority of family memers were non-hispanic white (86%) an female (68%). Approximately one-half were the patient s spouse (Tale 1 ). A total of 593 nurses (mean age, 42 years) returne at least one questionnaire. The meian numer of surveys complete per nurse was one (range, 1-10). The majority of nurses were non-hispanic white (83%) an female (86%). Tale 2 escries the family- an nurse-assesse outcomes an ocumentation of inicators of palliative care for the four physician specialties. Family-Assesse Outcomes In the ajuste analyses, only family memer ratings of quality of ying were significantly ifferent y physician specialty. Compare with patients care for y meicine attening physicians, family ratings were higher for patients with a neurology or neurosurgery attening physician ( Tale 3 ). There were no significant ifferences in family ratings of satisfaction with care in the ICU. Nurse-Assesse Outcomes In the ajuste analysis, two of the nurse-assesse outcomes were significantly ifferent across the four journal.pulications.chestnet.org CHEST / 145 / 2 / FEBRUARY

4 Tale 1 Characteristics of Patients an Families y Specialty of Attening Physician at Time of Death All Patients Meicine Surgery Neurology Neurosurgery Patients No. Statistic No. Statistic No. Statistic No. Statistic No. Statistic Age, mean (SD), y 3, (15.2) 2, (14.8) (17.1) (14.7) (18.5) Female 3, , Minority race/ethnicity 3, , Primary conition 3,124 2, Trauma Cancer Other Eucation 3,044 2, th grae Some high school High school grauate or equivalent Some college y college egree Postcollege stuy Ha insurance 3, , Family memers Age, mean (SD), y 1, (14.3) (14.6) (13.6) (14.4) (12.8) Female 1, Minority race/ethnicity 1, Patient s spouse 1, Data are given as No. (%) unless otherwise inicate. physician specialties ( Tale 4 ). Using meicine as the reference group, nurse ratings of quality of ying were significantly higher among patients care for y neurology or neurosurgery attening physicians an significantly lower among patients care for y surgery attening physicians. There were also significant ifferences across the physician specialties in nurse ratings of quality of care y all proviers. Although none of the iniviual physician specialties iffere significantly from meicine, when all specialties were inclue, incluing the negative rating for surgery specialty, the four specialties were significantly ifferent from one another. There were no specialty ifferences in nursing satisfaction with meeting family nees. Documentation of Palliative Care In the ajuste analyses, all ut one of the palliative care inicators were significantly ifferent across the physician specialties ( Tale 5 ). Using meicine as the reference, patients care for y neurology an neurosurgery attening physicians ha fewer palliative care consultations an fewer ocumente pain assessments, ut more avoiance of CPR, family conferences, an iscussions of prognosis in the first 72 h of their ICU stay. In aition, patients with a neurology attening physician ha more DNR orers in place at time of eath, spent fewer ays in the ICU, an ha shorter time to withrawal of mechanical ventilation. Using meicine as the reference group, patients with a surgery attening physician ha fewer pallia- tive care consultations, fewer DNR orers in place at time of eath, less withrawal of life-sustaining therapies, less avoiance of CPR prior to eath, fewer iscussions of prognosis in the first 72 h, more ays in the ICU, an longer time to withrawal of mechanical ventilation. Discussion We escrie several ifferences among patients care for y meicine, surgery, neurology, an neurosurgery attening physicians in the quality of EOL care in the ICU. We foun that patients who ha a neurology or neurosurgery attening physician at the time of eath ha higher family an nurse ratings of quality of ying than patients who ha a meicine attening physician, while patients with a surgery atten ing physician ha lower nurse ratings of quality of ying than patients who ha a meicine attening physician. Interestingly, there were no ifferences in family or nurse ratings of satisfaction with care when comparing these groups. These iscrepant finings etween the two outcomes (ie, quality of ying an satisfaction with care) may e explaine y ifferences in the experiences that these surveys measure. The Family Satisfaction in the ICU survey asks family memers to rate experiences with proviers, incluing the courtesy shown y staff, the type an completeness of information provie, an the help receive with ecision-making. The QODD questionnaire asks family memers to rate experiences that are irectly associate with ying. 316 Original Research

5 journal.pulications.chestnet.org CHEST / 145 / 2 / FEBRUARY Tale 2 Family- an Nurse-Assesse Outcomes an Documente Inicators of Palliative Care, y Specialty of Attening Physician All Patients Meicine Surgery Neurology Neurosurgery Outcomes No. Statistic No. Statistic No. Statistic No. Statistic No. Statistic Family-assesse outcomes Quality of ying a 1, (3.1) (3.1) (3.4) (2.0) (2.6) Satisfaction with care 1, (20.3) (20.3) (23.1) (17.3) (16.0) Satisfaction with ecision-making 1, (22.3) (22.0) (25.4) (20.4) (20.0) Total satisfaction 1, (20.3) (20.3) (22.9) (18.0) (16.3) Nurse-assesse outcomes Quality of ying rating a 1, (2.8) (2.8) (3.1) (1.9) (2.3) Quality of care y all proviers c 1, (1.6) (1.5) (1.8) (1.1) (1.6) Satisfaction, team met family s nees 1, (2.0) (2.0) (2.1) (1.9) (2.1) Documente inicators of palliative care, % Palliative care consultation 2, , Social work support 3, , Spiritual care 3, , DNR orer 3, , Life sustaining therapy withhel/withrawn 3, , Pain assessment 3, , CPR avoie in last hour 3, , Family conference 3, , Prognosis iscusse 3, , Days in ICU 3, (9.0) 2, (9.1) (10.0) (4.0) (5.9) Days to ventilator withrawal 1, (10.0) 1, (9.9) (10.5) (4.5) (5.7) Data are given as mean (SD) unless otherwise inicate. DNR 5 o not resuscitate. a Score coul range from 0 (terrile quality) to 10 (almost perfect quality). Score coul range from 0 (not satisfie at all) to 100 (very satisfie). c Score coul range from 0 (worst possile) to 10 (est possile).

6 Tale 3 Association of Attening Physician Specialty With Family-Assesse Outcomes a Regression Coefficient Family-Reporte Outcome No. P Value Meicine Surgery Neurosurgery Neurology Quality-of-ying rating c 1, Ref e Satisfaction with care f 1, Ref Satisfaction with ecision-making g 1, Ref Total satisfaction h 1, Ref Ref 5 reference. a Associations were teste with multipreictor linear regression moels with roust SEs, using a restricte maximum-likelihoo estimator. All moels inclue covariate ajustment for hospital (13 ummy inicators) in aition to outcome-specific confouner ajustments note in susequent tale footnotes. The overall P value for physician specialty was ase on the reuction in eviance otaine in a moel in which the coefficients for the three ummy inicators for physician specialty were freely estimate, when compare with a moel in which the three specialty-relate regression coefficients were constraine to 0.0. c Score coul range from 0 (terrile quality) to 10 (perfect quality). This moel inclue covariate ajustment for the family memer s age. P,.05. e P,.001. f Score coul range from 0 (not satisfie at all) to 100 (very satisfie). This moel inclue covariate ajustment for the patient s age, sex, eucation, an insurance status an the family memer s age an racial minority status. g Score coul range from 0 (not satisfie at all) to 100 (very satisfie). This moel inclue covariate ajustment for patient s age, eucation, an insurance status an the family memer s age. h Score coul range from 0 (not satisfie at all) to 100 (very satisfie). This moel inclue covariate ajustment for the patient s age, eucation, an insurance status an the family memer s age an racial minority status. Therefore, our finings likely reflect ifferences etween physician specialties in the ways in which family memers an nurses rate the patients experience of ying, while suggesting few ifferences in satisfaction with the critical care that is provie prior to eath. This stuy cannot ifferentiate the influence of physician specialty from the influence of the ifferent types of patients care for y these specialties. Neurology an neurosurgery patients likely have more acute, evastating injuries. Our ata show that patients with neurology or neurosurgery attening physicians have fewer ays in the ICU an less time to withrawal of mechanical ventilation than meicine patients, supporting this hypothesis. In aition, patients with neurology an neurosurgery attening physicians have fewer ocumente assessments of pain, which may also support our hypothesis that these patients have severe neurologic injury an may e unresponsive or comatose with little sensation of iscomfort or awareness. Also, there may e less prognostic uncertainty in cases of evastating neurologic injury that result in eath in the ICU. A prior stuy showe that Tale 4 Association of Attening Physician Specialty With Nurse-Assesse Outcomes a Regression Coefficient Nurse-Assesse Outcome Patients Nurses P Value Meicine Surgery Neurosurgery Neurology Quality-of-ying rating c 1, ,.001 Ref e f Quality of care y all proviers g 1, Ref Satisfaction, team met family s nees h 1, Ref See Tale 3 legen for expansion of areviation. a Associations for all outcomes were teste with complex multipreictor regression moels, with patients clustere uner nurses an estimates ase on restricte maximum likelihoo. Two outcomes (the quality of ying rating, satisfaction with how well the team met the family s nees) were teste with linear regression; the other outcome (quality of care y all proviers) was censore from aove an was teste with Toit regression. All moels inclue covariate ajustment for hospital (13 ummy inicators) in aition to outcome-specific confouner ajustments note in susequent tale footnotes. The overall P value for physician specialty is ase on the reuction in eviance otaine in a moel in which the coefficients for the three ummy inicators for physician specialty were freely estimate, when compare with a moel in which the three specialty-relate regression coefficients were constraine to 0.0. c Score coul range from 0 (terrile quality) to 10 (perfect quality). P,.05. e P,.001. f P,.01. g Score coul range from 0 (worst possile care) to 10 (est possile care). This moel inclue covariate ajustment for nurse s racial minority status. h Score coul range from 0 (not satisfie at all) to 10 (very satisfie). This moel inclue covariate ajustment for patient isease (cancer, trauma, other) an nurse s racial minority status. 318 Original Research

7 Tale 5 Association of Attening Physician Specialty With Meical-Recor-Assesse Outcomes a Regression Coefficient Meical Recor Outcome No. P Value Meicine Surgery Neurosurgery Neurology Palliative consult 2,193 c,.001 Ref e Social work services e 3, Ref Spiritual care f 3,121,.001 Ref e DNR in place g 3,110,.001 Ref e h LST withrawn/withhel i 3, Ref Pain assessment 3, Ref h CPR avoie, last hour 3,106,.001 Ref e h Family conference, first 72 h 3,109,.001 Ref h Prognosis iscusse, first 72 h j 3,107,.001 Ref h e Days in ICU 3,122,.001 Ref e e Time to MV withrawal j 1,581,.001 Ref e e LST 5 life-sustaining therapy; MV 5 mechanical ventilation. See Tale 2 an 3 legens for expansion of other areviations. a Associations for all outcomes except those relate to time (ays in ICU an time to MV withrawal) were teste with multipreictor logistic regression moels; the time-relate variales were teste with Cox moels (for Cox moel coefficients, the higher the value, the shorter the associate time perio). All estimates are ase on restricte maximum likelihoo. All moels inclue covariate ajustment for hospital (13 ummy inicators) in aition to outcome-specific confouner ajustments note in susequent tale footnotes. The overall P value for physician specialty is ase on the reuction in eviance otaine in a moel in which the coefficients for the three ummy inicators for physician specialty were freely estimate, when compare with a moel in which the three specialty-relate regression coefficients were constraine to 0.0. c From the initial 3,121 recors with vali ata on all preictors, 928 recors were not use in the coefficient estimates for palliative care consult ecause this outcome was uniformly 0 at five of the hospitals, an recors from those hospitals were roppe for purposes of estimation. P,.01. e P,.001. f This moel inclue covariate ajustment for patient age. g This moel inclue covariate ajustment for patient age an sex. h P,.05. i This moel inclue covariate ajustment for isease (cancer, trauma, other). j This moel inclue covariate ajustment for patient age an isease (cancer, trauma, other). patients with more severe neurologic injury an a iagnosis of suarachnoi hemorrhage or ischemic stroke were more likely to unergo withrawal of mechanical ventilation, suggesting that EOL ecisions in this population are often ase on the severity of the acute neurologic conition. 6 In aition to higher ratings of quality of ying, patients with a neurology or neurosurgery attening physician ha more chart ocumentation of some inicators of palliative care than patients with a meicine attening physician, while patients with a surgery attening ha fewer ocumente inicators. There were some exceptions to this pattern. For example, patients with a neurology or neurosurgery attening physician ha fewer ocumente pain assessments, which may reflect their overall neurologic conition. All other physician specialties patients ha fewer palliative care consultations than patients of meicine attening physicians. Palliative care consultations possily were not viewe to e necessary as often for the neurology an neurosurgery patients ecause of their shorter ICU lengths of stay, lower levels of consciousness, an, perhaps, less prognostic uncertainty. For surgical patients with less overall ocumentation of inicators of palliative care, increase use of palliative care consultation may represent a target for quality improvement. 5,27 Surgical patients an surgical practice may present unique challenges for integration of palliative care into the ICU. 5,28,29 The majority of eaths in the surgical ICU occur after a prolonge hospital course complicate y multiorgan failure with intermittent perios of improvement an eterioration These cases may provie unique challenges to proviing patients an families with prognostic information. In aition, the primary ethical principle governing care in the surgical ICU may e ifferent than that in a nonsurgical ICU, with more focus on a covenantal ethic rather than an ethic of scarce resources This stuy has several important limitations. First, there may e misclassification, with patients categorize ase on the specialty of the attening physician at time of eath. Furthermore, when categorizing patients y physician specialty, we are capturing ifferences oth in the types of patient care for y specific specialties, as well as ifferent types of patients. This stuy cannot aequately separate the influence of patients from health-care proviers, although the pattern foun with ifferences in quality of ying ut not satisfaction with care suggest that patient factors may play an important role. Secon, there may e other important, potentially confouning characteristics of the ICUs in this stuy, incluing staffing moels, journal.pulications.chestnet.org CHEST / 145 / 2 / FEBRUARY

8 multiisciplinary rouning, an nursing protocols, in aition to physician-level characteristics, that are not measure in our ata. Thir, while we suspect that patients with a neurology or neurosurgery attening physician ha lower levels of consciousness uring their ICU stay, we coul not confirm this suspicion. Glasgow Coma Scale was ocumente in only some of the hospitals, an even in those hospitals with regular ocumentation of Glasgow Coma Scale, it was only ocumente in a minority of patients in the last 24 h of life. Fourth, the response rates for the family an nurse surveys, while typical for this type of research, are low an may introuce nonresponse ias. 36 Fifth, severity of illness scores such as Simplifie Acute Physiology Score or APACHE (Acute Physiology an Chronic Health Evaluation) were not astracte for this stuy, as all patients ie uring the stuy. It is possile that this information woul help character ize the severity of illness at presentation an provie aitional insights to these results. Finally, while a strength of this stuy is that it was conucte at multiple hospitals, all are locate in a limite geographic region, which may limit our aility to generalize to other regions. Conclusions Family an nurse satisfaction with EOL care was not associate with ICU physician specialty, ut patients with neurology or neurosurgery attening physicians ha higher family an nurse ratings of quality of ying than patients with meicine attening physicians an a ifferent pattern of inicators of palliative care. Patients with surgery attening physicians ha lower nurse ratings of quality of ying an fewer ocumente inicators of palliative care. These finings may provie insights to improve the quality of ying for all patients. Interventions to provie quality EOL care in the ICU may nee to take attening physician specialty an patient iagnosis into account y targeting specific quality inicators or y aapting interventions to target patient an physician ifferences. Acknowlegments Author contriutions: Dr Kross ha full access to all of the ata in the stuy an takes responsiility for the integrity of the ata an the accuracy of the ata analysis. Dr Kross: contriute to the stuy concept an esign, ata analysis an interpretation, rafting an revision of the manuscript, an approval of the final version an serve as principal author. Dr Engelerg: contriute to the stuy concept an esign; ata collection, analysis, an interpretation; rafting an revision of the manuscript; an approval of the final version. Ms Downey: contriute to ata collection, interpretation, an analysis; revision of the manuscript; an approval of the final version. Dr Cuschieri: contriute to ata analysis, revision of the manuscript, an approval of the final version. Dr Hallman: contriute to ata analysis, revision of the manuscript, an approval of the final version. Dr Longstreth: contriute to ata analysis, revision of the manuscript, an approval of the final version. Dr Tirschwell: contriute to ata analysis, revision of the manuscript, an approval of the final version. Dr Curtis: contriute to the stuy concept an esign; ata collection, analysis, an interpretation; rafting an revision of the manuscript, an approval of the final version. Financial/nonfinancial isclosures: The authors have reporte to CHEST that no potential conflicts of interest exist with any companies/organizations whose proucts or services may e iscusse in this article. Role of sponsors: The sponsors ha no role in the esign of the stuy, the collection an analysis of the ata, or the preparation of the manuscript. Aitional information: The e-appenix can e foun in the Supplemental Materials area of the online article. References 1. Angus DC, Barnato AE, Line-Zwirle WT, et al ; Roert Woo Johnson Founation ICU En-Of-Life Peer Group. Use of intensive care at the en of life in the Unite States: an epiemiologic stuy. Crit Care Me ;32(3): Teno JM, Gozalo PL, Bynum JPW, et al. Change in en-of-life care for Meicare eneficiaries: site of eath, place of care, an health care transitions in 2000, 2005, an JAMA ;309(5): Prenergast TJ, Claessens MT, Luce JM. A national survey of en-of-life care for critically ill patients. Am J Respir Crit Care Me ;158(4): Baggs JG, Norton SA, Schmitt MH, Domeck MT, Sellers CR, Quinn JR. Intensive care unit cultures an en-of-life ecision making. J Crit Care ;22(2): Mosenthal AC, Weissman DE, Curtis JR, et al. Integrating palliative care in the surgical an trauma intensive care unit: a report from the Improving Palliative Care in the Intensive Care Unit (IPAL-ICU) Project Avisory Boar an the Center to Avance Palliative Care. Crit Care Me ; 40 ( 4 ): Diringer MN, Ewars DF, Aiyagari V, Hollingsworth H. Factors associate with withrawal of mechanical ventilation in a neurology/neurosurgery intensive care unit. Crit Care Me ;29(9): Schaller C, Kessler M. On the ifficulty of neurosurgical en of life ecisions. J Me Ethics ;32(2): Gujjar AR, Deiert E, Manno EM, Duff S, Diringer MN. Mechanical ventilation for ischemic stroke an intracereral hemorrhage: inications, timing, an outcome. Neurology ;51(2): Hoe NM, Engelerg RA, Treece PD, Steinerg KP, Curtis JR. Factors associate with nurse assessment of the quality of ying an eath in the intensive care unit. Crit Care Me ;32(8): Curtis JR, Treece PD, Nielsen EL, et al. Integrating palliative an critical care: evaluation of a quality-improvement intervention. Am J Respir Crit Care Me ;178(3): Curtis JR, Nielsen EL, Treece PD, et al. Effect of a qualityimprovement intervention on en-of-life care in the intensive care unit: a ranomize trial. Am J Respir Crit Care Me ;183(3): Treece PD, Engelerg RA, Shannon SE, et al. Integrating palliative an critical care: escription of an intervention. Crit Care Me ;34(suppl 11 ):S380-S Dillman DA. Mail an Internet Surveys: The Tailore Design Metho. New York, NY: John Wiley & Sons, Inc ; Curtis JR, Patrick DL, Engelerg RA, Norris K, Asp C, Byock I. A measure of the quality of ying an eath. Initial valiation 320 Original Research

9 using after-eath interviews with family memers. J Pain Symptom Manage ;24 (1 ): Levy CR, Ely EW, Payne K, Engelerg RA, Patrick DL, Curtis JR. Quality of ying an eath in two meical ICUs: perceptions of family an clinicians. Chest ;127 (5 ): Mularski R, Curtis JR, Osorne M, Engelerg RA, Ganzini L. Agreement among family memers in their assessment of the Quality of Dying an Death. J Pain Symptom Manage ;28 (4 ): Glavan BJ, Engelerg RA, Downey L, Curtis JR. Using the meical recor to evaluate the quality of en-of-life care in the intensive care unit. Crit Care Me ;36 (4 ): Heylan DK, Tranmer JE ; Kingston General Hospital ICU Research Working Group. Measuring family satisfaction with care in the intensive care unit: the evelopment of a questionnaire an preliminary results. J Crit Care ;16 (4 ): Heylan DK, Rocker GM, Doek PM, et al. Family satisfaction with care in the intensive care unit: results of a multiple center stuy. Crit Care Me ;30 (7 ): Wall RJ, Engelerg RA, Downey L, Heylan DK, Curtis JR. Refinement, scoring, an valiation of the Family Satisfaction in the Intensive Care Unit (FS-ICU) survey. Crit Care Me ;35 (1 ): Downey L, Engelerg RA, Shannon SE, Curtis JR. Measuring intensive care nurses perspectives on family-centere enof-life care: evaluation of 3 questionnaires. Am J Crit Care ;15 (6 ): Clarke EB, Curtis JR, Luce JM, et al ; Roert Woo Johnson Founation Critical Care En-Of-Life Peer Workgroup Memers. Quality inicators for en-of-life care in the intensive care unit. Crit Care Me ;31 (9 ): Mularski RA, Curtis JR, Billings JA, et al. Propose quality measures for palliative care in the critically ill: a consensus from the Roert Woo Johnson Founation Critical Care Workgroup. Crit Care Me ;34 (suppl 11 ):S404-S Gries CJ, Curtis JR, Wall RJ, Engelerg RA. Family memer satisfaction with en-of-life ecision making in the ICU. Chest ;133 (3 ): Kross EK, Engelerg RA, Gries CJ, Nielsen EL, Zatzick D, Curtis JR. ICU care associate with symptoms of epression an posttraumatic stress isorer among family memers of patients who ie in the ICU. Chest ;139 (4 ): Muni S, Engelerg RA, Treece PD, Dotolo D, Curtis JR. The influence of race/ethnicity an socioeconomic status on enof-life care in the ICU. Chest ;139 (5 ): Braley CT, Brasel KJ. Developing guielines that ientify patients who woul enefit from palliative care services in the surgical intensive care unit. Crit Care Me ;37 (3 ): Aslakson RA, Wyskiel R, Thornton I, et al. Nurse-perceive arriers to effective communication regaring prognosis an optimal en-of-life care for surgical ICU patients: a qualitative exploration. J Palliat Me ;15 (8 ): Aslakson RA, Wyskiel R, Shaeffer D, et al. Surgical intensive care unit clinician estimates of the aequacy of communication regaring patient prognosis. Crit Care ;14 (6 ):R Lauplan KB, Kirkpatrick AW, Korteek JB, Zuege DJ. Longterm mortality outcome associate with prolonge amission to the ICU. Chest ;129 (4 ): Lipsett PA, Swooa SM, Dickerson J, et al. Survival an functional outcome after prolonge intensive care unit stay. Ann Surg ;231 (2 ): Mayr VD, Dünser MW, Greil V, et al. Causes of eath an eterminants of outcome in critically ill patients. Crit Care ;10 (6 ):R Cassell J, Buchman TG, Streat S, Stewart RM. Surgeons, intensivists, an the covenant of care: aministrative moels an values affecting care at the en of life Upate. Crit Care Me ;31 (5 ): Schwarze ML, Remann AJ, Alexaner GC, Brasel KJ. Surgeons expect patients to uy-in to postoperative life support preoperatively: results of a national survey. Crit Care Me ; 41 (1 ): Schwarze ML, Braley CT, Brasel KJ. Surgical uy-in : the contractual relationship etween surgeons an patients that influences ecisions regaring life-supporting therapy. Crit Care Me ;38 (3 ): Kross EK, Engelerg RA, Shannon SE, Curtis JR. Potential for response ias in family surveys aout en-of-life care in the ICU. Chest ;136 (6 ): journal.pulications.chestnet.org CHEST / 145 / 2 / FEBRUARY