Portland State UniversityPDXScholar

Sociology Faculty Publications and Presentations Sociology

8-2003

Next-of-Kin Perceptions of Physician Responsiveness toSymptoms of Hospitalized Patients Near DeathJoel C. CantorRutgers University - New Brunswick/Piscataway

Jan BlusteinNew York University

Matthew J. CarlsonPortland State University, carlsonm@pdx.edu

David A. Gould

Let us know how access to this document benefits you.Follow this and additional works at: http://pdxscholar.library.pdx.edu/soc_fac

Part of the Health Services Research Commons, and the Medicine and Health Commons

This Article is brought to you for free and open access. It has been accepted for inclusion in Sociology Faculty Publications and Presentations by anauthorized administrator of PDXScholar. For more information, please contact pdxscholar@pdx.edu.

Citation DetailsJoel C. Cantor, Jan Blustein, Mathew J. Carlson, and David A. Gould. Journal of Palliative Medicine. August 2003, 6(4): 531-541.

or* I'ALiJAi iv rVnlimii- h. Number 4, 2()l).1(0 Marv Ann l.ifber!. Inc.

Next-of-Kin Perceptions of Physician Responsiveness toSymptoms of Hospitalized Patients Near Death

C. CANTOR, ScD.' , IAN I5LUSTHIN, M.D., Pb.D.,- MATHI-VV j . CARLSON, Ph.D./and DAVID A. GOUI.D, Ph.D.'

ABSTRACT

Many different medical providers visit critically ill patients during a hospitalization, and pa-tients and family members may not feel any physician is truly in charge of care. This studyexplores whether perceiving that a physician was clearly in charge is associated with reportsby surviving next of kin about the responsiveness of physicians to symptoms in hospitalizedpatients near the end of life. We conducted telephone interviews with surviving next of kinof adult patients (// = 1107) who died in one of five New York City teaching hospitals be-tween April 1998 and June 1999 after a minimum 3-day inpatient stay. Next-of-kin ratings ofwhether physicians did "all they could" all or most of the time in response to patient pain,dyspnea, and affective distress (confusion, depression or emotional distress) were comparedby whether the next of kin reported one or more physicians "clearly in charge" of care, ad-justing for patient and next-of-kin characteristics. More than 80% of patients were reportedto have experienced often serious pain, dyspnea, or affective distress. Physicians were ratedas responsive to pain by 79.1% of respondents, to dyspnea by 84.9%, and to affective distressby 66.6"o. Ratings of physician responsiveness to pain (/; = 0.001) and affective distress ip =0.001) were significantly lower among patients for whom no physician was seen as clearly incharge of care. This finding is consistent with the view that ensuring that a physician coor-dinates the care of seriously ill, hospitalized patients may improve symptom management.Further research is warranted to establish causality and identify optimal models of care.

INTRODUCTION fienls dying in acute care hospitals had moderateto severe pain at least half of the timedLU'injj, their

F OK I'MMNrs NLAK \iu- h\i'Ol' I 111, palliation of final days of life.' Other studies confirm thai the

symptoms is ot" paramount impoiiance. Yet tailuro to control sympkims in hospitali/od dy-studies sufjjgest that those spending their final ing patients is widespread.-'-'days in hospitals often die with symptoms This failure is significant because more thanunchecked. I'or example, in the Study t(t Under- one half of deaths in the United States occur instand Pr(\gni>sis and Prefcivncos for Outcomes hospitals.^ I lospitals aro complex institutions inand Risks o\ Treatment (SUPPORT), half of pa- which providers witb ditferent roles and por-

'Ci'nk-r lor State l ieal lh Pol ic \ , Rutgers, the State University of New \cv^v\, New nriinswii-k, New lei'soy.-Wai^ner tii'(Kluate School, New >'ork University. New ^o rk . \e^^• York.'IX'[MrtineiU ot Sociologv, Portland State Uni\ ei-sitv, .md ()rc't;o)i He.iith ("i: Science L 'n i \e r s i i \ , PortkinLJ, i,")rr>;on.

I ! iosjiii.ll i n n d . New York, New \\>rk.

531

532 CANTOR ET AL.

spectives share responsibility for patient care.Sometimes the houndaries of responsibility andaccountability for patient cnro are unclear or evencontentious.'' These ambiguities are often inten-sified in the care of seriously ill patients,'' in in-tensive care settings,''-^ and in academic medicalcenters."' Moreover, during hospitalization, pro-viders who have not previt)usly been invt'iK'ed inthe care of the patient—including subspecialists,consultants, and hospitalists—sometimes take oncentral roles in patient care. Some have suggestedthat Ihis unfamiliarity breeds discontinuity,which in turn undermines the quality of hospital

care.How might ambiguities and discontinuities of

responsibility be related to the quality of care thatis delivered to hospitalized patients at the end oflife? We addressed this question by examiningthe relationship between next-of-kin reports ofwhether one or more physicians were "clearly incharge" of patient care and whether the pliysi-cians did "all they could" to control symptomsamong patients who died in five New York Cityteaching hospitals.

The perspectives of surviving next of kin art-important in their own right and are i^elevantmarkers of the quality of care. Next of kin are im-portant participants in and observers ot caru.Fainily members often spend a groat deal moretime in bospitals with patients than do physiciansor the hospital staff, and they are frequently pres-ent for physician-patient discussions or serve asproxy decision-makers for inconipetont patients.Prior studies have shown that surrogates such asfamily members are moderately accurate raters ofsymptoms such as pain, although they may over-report symptoms and may be more critical thanpatients of caregivcrs.'""'-^

As clinicians .strive to improve symptom man-agement in patients near death, il is important tounderstand tho correlates of good outcomes. Thisknowlodgo can inform the dosign of care strato-gios, such as special palliative caro units or pal-liativo care consultation teams, as well as guidethe design of future research in this area.

METHODS

setting and sample

The fivo study hospitals woro participants in aprogram to improve tho care ot patients near tho

ond ot' life sponsored by tho United Hospital|-Linci of Now York.'"' Tho hospitals rango from51 I ti> 1027 bods in sorvice, and havo botvveen 35and 74 modical residents por 100 beds (authors'tabulations of 1998 Now York State InstitutionalCost Reports). Two ot" the study sites are aca-demic hoalHi centers affiliatod with medicalschoiils.

A probability sample of 2528 patients 18 yearsof ago or older with a minimum stay of 3 daysprior to dying between April 1998 and June 1999was drawn from the hospitals' computorizedrocords. Patients rocoiving special palliative caresorvicos wore oversamplod in two oi tho hospi-tals. We use statistical controls to address the pos-sibility of bias ijitroducod by Hio interventionovorsample (see Analysis below).

The Institutional Review Boards of the tivo par-ticipating hospitals approvod the research re-ported in this paper. Beginning in November1998, lottors describing tho study and stating thatresponses woro voluntary and confidontial weremailed no sooner than 6 wooks after death to sur-viving next of kill listed in hospital records. In-terviewing was conductod between Docomber1998 and November 1999.

Study dala

The primary data ior the study were collectedduring computer-assistod telephone interviewswith the Hstod next of kin botvveen 2 and 17months after the pationt's death. Half of inter-viows were completed botwoon 5 and 7 monthsattor death and 99''n were complotod within 12months aftor death. Interx'iews avoraged 22minutes and woro conducted by oxperiencod in-torviewors who had boon trained to administertho study questionnaire and to consider tho spe-cial noods o\- boroavod and oldorly respondonts.Atter obtaining oral intbrmod consent, inter-viewers askod respondonts about porceptions oftho patients' conditions and care during the hos-pital stay during which tho palionts diod. In ad-dition, data iTom computori/od hospital rocordson longth of stay, oxpectod primary source ofpaytnont, admitting diagnosis, and patient agoand gonder woro linked to the data derivedIrom tho telephone intorviows. Although ot po-tential importance, data on the characteristics otphysicians or other oaro providors were un-available.

Tbe outcomes examined in this papor are basod

PHYSICIAN RESPONSE TO SYMPTOMS IN DYING PATIENTS 533

on respondent assessmejits of physicians' lovol ofeffort to treat patient symptonis. The respondontswore asked about the extent to which patients ox-porioncod pain, shortness of breath, and affectivedistress (confusion, depression or emotional dis-tress). Thon, thoso roporting that the patient ox-porioncod symptom(s) were asked for theirassessment of tho adoquacy of physician respon-sivonoss to symptoms. For example, for patientswho oxporienced pain, resp(nidonts wore asked,"Do you think that the doctors did everythingthey couid to help cotitriil (patient's name/rela-tionship) pain . . . all of the time, niost of the timo,about half ot the timo, somo of the time, or nonoof tho timo?" Respondents woro askod parallelquestions aboLtt physicians' eftoi"ts to help pa-tients "broatho more easily" and to relieve "con-fusion, depression or omotional distress."

hi a sopai'ate series ot questions, respondentswero asked about tho physicians caring for thopationts dLu-ing tho hospital stay. First, respon-dojits wore asked, " . . . was there one doctor whowas clearly in charge ot (patient's namo/ rela-tionship) caro, more tban ono doctor in chargo, orno doctor who was clearly in charge of (his/her)care?" "Ihoso who reported that there was ono ormoro dtx'tors "clearly in charge" were thon askedwhether thoso physicians wero " . . . involved incaring for (patient's name/relationship) beforethat hospital stay?"

Data woro also colloctod to control for factorsthat might confound tho association betweenrospondont roports of whothor thero was aphysician clearly in charge and tho physicianrosponsivenoss to symptoms. Covariatos in-cluded rospondent and patient demographicand sociooconomic characteristics as well as pa-tiont living arrangements and health and dis-ability status prior to admission. Additionally,because prior rosoarcb has shown respondentexpectations to be associated with satisfactionwith care, respondents wero asked when theytirst realized that the pationt would dio.'" '̂"^Somo rospondonts did not know or refused toanswer some survoy questions and somo datawero missing trom hospital administrativerocords. Casos with missing valuos were ex-cluded irom analysis, oxcept for household in-como in tho last year and major diagnostic cat-egory whoro more than 10')'.. of tho cases woromissing. For thoso variables, to avoid potontialsoloctitm bias a "missing" category was in-cluded in tho analvsis.

Anali/sis

Aftor tabulating sample characteristics, wo ox-aminod tho association between tho roportodle\'ol oi physician efforts to addross symptoniband rospondent and pationt charactoristics. Wofocused on tho association of vvhethor thoro wasa physician(s) cleariy in charge of patiojitcai-o andthe rosponsivonoss of physicians to symplorns.Rospondont and patient covariatos thai: wo toundto be signiiicantly associated with physician re-sponsiveness in bivarialo ,v- tosts at a levol of /' <0.05 or iowor woro controlled lor in estimatingadjusted ocids ratios in logistic rogrossion mod-ols. In addition, to help ensure that estimates werounbiased, tho rogrossion models also controlledfor tho sampling strata (i.o., hospital and pallia-tive care program participation). As woll, unad-justed odds ratios for physician-in-chargo \'ari-ables woro computed for comparison to {hoadJListod ratios for thoso \'ariables.

RESULTS

Of the initial 2528 caso>s, 209 (8.3".,) hospitairocords did not have adequate contact informa-tion for a survix'ing noxt of kin and 61 (2.4',*n) hadcontact information, bitt tho named porstin ro-portod ha\ ing had no interaction with tho hospi-tal statf caring for the pationt. Theso casos woroconsidorod ineligible for the study. Of tho re-maining 2258 casos, 412 (18.27o) could not bo lo-catod using the contact information provided; 176(7.8"'..) could not be reached after nutitiplo at-tempts, 58 (2.6"''n) woro unable to participate bo-causo of incapacitation or language barrier, and350 (15.5"<0 rofiisod to participate. lntor\'iowsworo conducled with the remaining 1271 noxt ofkin, yiolding a responso rate of 56.3"''n. Whilo lossthan ideal, this rale is comparable to studios ofpatient oxpcriencos and satisfaction roportod intho litoraturo.'*'-'"

Rosponso ratos varied among tho fivo h(ispitalsfrom 46.3"'';. to 63.3';'a (/; < 0.001); and wero Ioworfor pationts with an expected payment souroo ofModicaid or uninsured (47.0"^) oomparod toMedicare (57.2';''o) or privately insurance (64.9"'o;/' = O.OOl) and for women patients (54.3"'o) com-pared to tnon (58.5'K.; /) = 0.044). Rosponse ralosdid not vary significantly by pationt a^c, majordiagnostic catogory, or length of stay. Tho analy-sis presentod hore is limitod to 1 107 (87.1'%.) cases

534 CANTOR ET AL.

ratod by interviewers as knowledgeable on "all"or "most" questions on a four-point scalo (com-pared to "somo" or "very fow or nono" of thoqitostions).

Tablo 1 doscribos tiio analysis samplo. With re-spoct to tbe main indopendent variables of intor-ost in this study, no doctor was soon as dearly inchargo in nearly 2O'''.i of casos, and approximatelyhalf of tho romaining respondents reportod thatalthough a physioian(s) was in chargo, he or shohad no relationship to the pationt prior to tho fi-nal hospitalization.

Respondonts were predominantly vvomonand prod(.>minantly nonolderly. Forty porcont re-portod that thoy had no moro than a high schooleducation, and the modal relationships to thopatient woro child and spouso/partnor. Morothan half of all rospondonts had nol oxpoclodthat tho patient would dio during the hospitalstay until near tho vory end. Howovor, a sizableminority, nearly a quarter, said that they roal-i/,od that the patient would dio prior to or justafter admission.

As shown in tho socond part of Tablo 1, studypationts were quito old and sick, with high pro-portions roportod in fair or poor boalth statLis orli\'ing in a nursing homo two months prior to ad-mission. Approximately half of the pationts worewomen, approximately half woro non-white, andmost reportedly spoko English "vory woll." Onavorago, the patients had lower educational at-tainment than thoir noxt of kin {p <•. 0.001) did,with only one third having moro than a highschool dogroo. Data from Ihe two thirds of re-spttndents who reported pationi: incomo showthat tho patients wero typically of modost moans.Modicaro was the predominant oxpected primarypayer, followod by pri\'ato covorago and Medi-caid or nt) coverage. Finally, the pationt popula-tion had a di\ orsity of admitting diagnoses, anda largo proportion bad lengthy hospital stays. Ro-ports of having ono or moro physician(s) "clearlyin chargo" varied by respondent and patient so-ciooctinomic circumstances, patient ]i\'ing situa-tion prior to admission and diagnosis; but not byother sampie charactoristics (data not shown).

As summarized in Table 2, more than 8 in 10pationts woro roportod to have oxporioncod pain,dyspnoa, or affective distress (confusion, depres-sion, ov emotitmal distress) during Iheir finalhospitalization. Respondont reports oi physicianrosp(.)nsivonoss \'ariod by symptom. Physician ro-spc)nsi\'onoss was ratod highly (detinod as ix'ports

oi physicians doing "all that thoy could" to ad-dross symptoms oither "all of the timo" or "mostof tho timo") in 79.1",, of pain casos, 84.9''u of dys-pnoa cases, and 66.6"'o of casos with affectivo dis-tl'OSS.

Physician rosponsi\'onoss was rated higherwhen a physician was seen as cloarly in chargo,although this difforonco was not significant fordyspnoa (Tablo 3). In tho caso ot pain, physicianrosponsivonoss ratings woro liighor v\'hen thophysician(s) in chargo had a rolationsiiip t:o thopatient prior to the hospitalization compared towhen no prior relationship was roportod (/'0.043).

So\"oral rospondont charactoristics wore also as-sociated with ratings of physician rosponsivo-ness. Older respondonts tondod to rate physiciansas boing moro responsive, as did rospondontswho had anticipated tho patient's doath prior totho admission. Few pationt factors wero associ-atod witb ratings of physician rosponsivoness.Notably, only prior health status was consistentlylinked with responsiveness. Highor physicianrosponsix'oness to symptcims assi»ciated withpoorer patient health. Consistent with the healthstatus finding, physician rosponsivonoss ratingswero highor ior patients who prox'iously livod inan institution such as a nursing home, allhoughthis was significant only for dyspnea. Other pa-tiont charactoristics (racc/ethnicity, oducation,English-spoaking ability, household income, ox-pectod primary payer class) as vvoil as oxpoctodpayer, diagnosis, longth of stay and bospit-al woronot consistontly associatod with physician re-sponsiveness tc) symptoms.

Logistic I'egression models aro ctinsistont withtho rosults shown above and demonstrate that thoassociation botwoon tho physician-in-charge \'ari-ablo and symptom managoment are not altorodv\'hen contrtils for co\'ariatos aro addod. As shownin Tabio 4, roports of ha\'ing a physician clearlyin charge was associatod with a groator odds ofa high rating of physician offort to addross painand al fee tivo distress com pa rod to ha\ ing nophysician in chargo in both unadjustod and ad-justed models. In tho caso of pain, significantlyhighor physician rosponsivonoss was found forpatients with a pbysician(s) in charge who caredfor thom prior to tho hospitalization comparedpationts with a physician(s) in charge but with-out a prior rolationship in tho LinadJListod model(odds ratio lORl == 1.6; confidence intorva! jCll =1.1, 2.5) although this association is of bordoriino

PHYSICIAN RESPONSE TO SYMPTOMS IN DYING PATIENTS 535

r \ i ; [ i 1, S A M I ' I I C ' I I A K A ( iFKi^i i rs

Physician "dearly iii charge" {ii - 1,1)93)None " 18.3>'cs, without prior relationship 39.1Yes, with prior relationship 42.3

RespondentWomen (;/ - 1107) 7(1.1Age griiup ill = I0S5)

14-39 yea r s I 7.<S411—1.4"' •>! 7

75 or older ,S.(1L d u c a t i o n {ii - I 1(12)

l.,ess than high school 12.3High school graduate 27.6Some college 14.2College graduate 40.*^*

l.iveLl w i th pa t i en t {it - I 104) 44.3Relationship to patient (;; - 1107)

SpoLise/partner 2(i. IParent 4.2Child 44.2SIhlinj; 9,(->Other ib.O

When death was First expected (;; - lilS2)i'rior to ^idmission •-).Hjust after admission 14.3.Abiuit lialF way through hospital stay 22.9Not e\pected or nnt sure ^3.0

PatientWomen {it - I 107) 30,1Age group, years {ii " 1107)

30 dnd undei" 1 3.831-64 IN.8'i3-74 I').273-8-i 26.4S3 or oider 14.3

Race/ethnicitv (/; - 1088)White, non-Hispanic 31,3Black, non-i Jispanic 27.2Hispanic Ih.^Other 4,g

I'ducation iu - 103(-.)Less than high school 33..jHigh school graduate 34.3Some cnllege 12.0Cnllege graduate 2(1.4

Spnke English "\'cry well" {n - lt)06) N6.7Healtii status—-2 months prinr tt) cKimission (;; - I0S2)

Kxceilent, gnnd, nr \'erv' j^tHui 2'-'.4l-''Tii- ' 30.0Poor 40,1

l.i\ed in institution—2 months prior to admission (// - !097) I'̂ .9Hinusehnld incnme List year (// -= 7'̂ 2)

Under $15,000 " 41.9$l5,tK)0-$49,999 3S.9$.St),llOO or mnie 22.1

Expected primary payer class {it - 1063)Medicare dl.dVk'dicaid, selt-pav, or charity care 15.'iPri\'ate coveratje "̂ 2 4

536 CANTOR ET AL.

TAISI !-• 1. S.win r CHAkAr"H"RisrK"s

Major di.ignnstic category {it • 947)NeophismsCirculatory systemRespiraiory systemDigesti\"e systeminFectious/parasiht. condititmsOther

Length of stay, days {ii - 1107)3-5(1 -9

10-1516-272S-more

Hospital (;; - 1107)

BCi)E

19.720.617.3').7

12.020.6

I4.,S21.622.t)20.121.6

18,3iS.821.321.320.2

significance in the adjusted tnodel (OR '-^ 1.6;CI = 1.11, 2.5). Having a prior relationship did nothave a significant association with physician re-sponsiveness for the other two symptoms (datanot sht>wn).

DISCUSSION

Consistent with prior studies of hospitalizedpatients near the end of life, respondents in OIH'study reported high levels oi symptoms in a se-ries of patients who died in the study hospitals.' '''

More than .S(.)"o of patients reportedly experi-enced pain, dyspnea, or affecti\'e distress. More-over, next-of-kin ratings of the adec-juacy of physi-cian efforts lo address the symptoms were notconsistently posilive, with between I5.2''vi and.13.5'''n reporting that physicians failed lo do "allthoy could" most oi the time to address symp-toms, depending on the symptom addressed.

Nearly 1 in 5 respondents reported that nophysician was clearly in charge of patient care,and these respondents rated physician respon-siveness to affective distress and pdin substan-tiiilly lower than respondents who reported that

TAisi 1 2 . N i M - o i - K i \ i Ki i v ^ K i i n S Y M P I I >vis

W l Ri P M « i\\\:\> r o n i . ' " A i i ' I ' m ^ C i

Svmptoin pri'valence

Oi patients with the symptom, reports oi' "doctorsdid al! tliev could" to address svniptoms, ".. distributi(*nAil oi tho timeMosi of the limeAbout half oi the timeSome of the timeNone of the time11

\\\> L)l I ,\<} 1 U

[(.1 A l l D R I

I'liin

82.4434

46,4.32,7

h.6

10.93.4

731

- W i i i n i

•SS S\MI 'rin'sici-w^

lOMs

Si/iiij'loiii

85.711)23

29.04.47.33.1'

838

80.4921)

37.:>24.1

8.414.'-!10,6

f.67

•'Confusion, depression or emi>tional distress.

PHYSICIAN RESPONSE TO SYMPTOMS IN DYING PATIENTS 537

• 3 . P i - [ ' ; t ; [ - . M (">i N l x ' l - c n - K i N K f i i i K i i M i . I I I A I P I I I ' S I C I . A N S D I I ' " A I I T i i i - i C o i i i n "

i ( > A D I I K I S S S v M i ' i D M s A M c i i < M O S I o r i i i i T i M i i i i S A M I ' I l C i I A K A L T I i ^ i s i i c s

chari!clcrisiic>

TntalPiiysician "clearly in charge"

NoneYes, without prior relationshipYes, with prior relationship

Respondent(ieniler

WomenMen

Ago gruiipI4-.34 years4t]-44 "'

h5-7475 or older

EducationLess than high sclioolNigh .school graduateSome cc'llegeCollege i;ra(.lu.itc'

I .ivei-l with patientYesNo

Relationship to patientSpoLise/partnei-ParentChildSiblingOther

When death was first expectedPrior to admissinnJust aftei' admissionAhout hall way thriuigh hospital stayNot oxpoctcd or not sui-e

PatientCieruler

WomenMen

Age group, yoars30 aiui under51 (-,465-747-V^4ST or older

Ra<,e/i.'thnicit\'White, non-1 lispanicBlack, non-Hispanii,'f iispann"Other

lklucafionLess than high scho<ilHigti school gi-aduateSiMne colk'geCollege graduate

Pain(n - 7311

79.1

(-(6.478.284.7

80.378.(1

72.175.1)84. h81.681.3

77.474.381,881,1

74. H78.1)

83.079.475384.674.3

90.5S7.I78.774.2

83.0

83.275.273 3

7(->.7S7.6

81).48t). I73,371.8

76,577.485.180.4

O.ODI

0.606

0.025

0.250

0.236

0.1)02

0.008

0,050

0.217

0.3.S0

M.S

86.186.

83.h

78.3S3.9S8.288.486.2

77.184.186.486.7

81.187.5

S7.580.082. S40.284.5

84.3

S0.5

82.787,1

S3.282.086.185.686.4

88.5S4.i78.S75.7

84, h84.784,0S(>.S

S3SI

0.118

0.522

o.o.s:

0.120

0.011

0.3 14

0.01)1

0.078

0.716

0.012

0.416

(11 - ('(••7)

Stl.468.171.7

bh.4b7.0

62.273.8

60.7

66,366.8

6h.480.76O.tl79.372.(.

80.083.366.4

67.465.8

68.362.6n4.263.4

67,170.063.153.6

M.268.766.75-4.2

P

O.0t)l

0.003

0.051

0.407

O.OO.S

0.00!

0.662

0,746

0.2S4

t).O24

538 CANTOR ET AL.

T A I I I . I 3 . P i K i . i : x i i.>i N i X I - C M - K I X R i T H K I I M . I I I A I P I I \ N | ( I . W S D I D " A I I T : I I " ^ C o m / '

\ o A i i i i K i ' S S S ^ A i i > n ' M s A l I O k M o ^ r o i n i l " T I M i i ; > ' S A M P I I- C i i \ I < A ; I I K I S T I C - ( C O \ I ' l ) ]

I'ltiii nu^pinui /),in - 7M) (n - S'iS) in

SivtipIc

Spoke English "very well" 11-918 0.188 0.264Yes

! Icaltii status 2 mnnths pnor U> admissionI'xcelleni, gond or \ e ry goodI'airPoor

l.i\'fd in inslilution 2 months prior to admissionYes.\]n

Hnusehnid income last yearUnder $15,000$15,(50(1-44,994S50,110(1 or moreUnknown/refused

lixpectei.1 primary payer classMedicareMedicaid, self-pay, Lir charity carePri\atL' co\erage

Major diagnostic categoryNeoplasmsCirculatory syskMnRespiratory systemDigesti\"e systemInfectious/parasilic conditionsOtherMissing

Length nf stay, days3 to 56 to 410 to 1516 tn 2728 nr iin>re

flnspital/\HCt:)E

S(1.280.7

72.175.486.4

86,177.8

82.376.774.1

78.0

74.1

77.278.1

74,277.181.177.384,472.084.4

75.3S2,9

77,682.875.7

69.782. S

76.581.5

83.1

(l,(!01

0.055

0,5520.391'

(1.S44

0.2640.343''

(1.301

0.031

85.481.0

7S.583.340.8

42.583.2

S8.282,784.283.8

85.580.284.7

8(1.487.483,488.1

93.478.186,0

83.584.086.286.683,6

77.486,284.784.0

86.2

(1.0(11

(1.006

()..384

0.25.'̂ '-

O..'̂ 26

(1.035() .02l ' '

0.888

0.(150

47.462.0

54,462.673.8

64.666.1

66.764.861.170.3

65,370.067,1

65.064.466.364.577.361.467.0

64.864.768,(170.464,5

6(1.4

67.265.064.464.(1

(1.002

0.55!

0.4170.642'

0.661

0.4400.322^'

0.800

0.615

•'Sample sizes for individual \ ariables may be less than the totals due to missing values, p valuos arc shown withand v\i{hout missing yalues for \ariahles vyitli IO'\. nr more missing,

'Xonfusinn, depressinn nr emotional distress.'!> value excluding " u n k n o w n / r e i u s e d " cases, pain [n - 445), dyspnea in - 554), affective distress (;; - 4.55).^'i' value excluding missing cases, pain (;/ - 6l6), dyspnea {ii ^ 717), afTccti\e distress {u =- 564).

a physici<Tn(s) was in charge. This contrast was citin efforts when a physician was seen as in•greatest ior affectivo distress symptoms, only half charge and he/she cared lor the patient prior toof respondents reported that physicians did "all the hospitnlizotion. Ct)mparable ratings of physi-they couid" all or most of the time to address dis- cian responsiveness to pain were 6fi.4''n with notress when no physician was seen as in charge, physician in charge and 84.7% when a physicianbut nearly three fourths reported adequate physi- with a prior relationship was in charge. Smaller

PHYSICIAN RESPONSE TO SYMPTOMS IN DYING PATIENTS 539

Ai i i I 4 . O i i D S K A I K ' S O l N " i - \ i - i i i " - K i N R I ' I ' O K H M ; I H \ I P I H S I C ' I A N S D I D

" A l l T i i r \ C o i n I ) " l o A D M K I s'^ S ^ M ^ T O V l ' - . A i i o i - ; M o s i o i - i l l i T I M I -

/s Riilio

I'liiit (n - Dyspnea In - 77il) r^<-' (n - 54:

ljiiiuiiti<lcii

Physician "clearly in charge"VVitii prinr re la t ionship ' 3,(1 (1,4, 4.4) 3,1 (1.8, 5.1) 1.6 (1.0, 2.7) 1.6 (0.4, 2.7} 2.4 (1.5, 3,8] 2.2 (1.3, 3.h)Wilhnut prinr ielati(»n.-,hip I.S (I . I , 2.4) 1.9 (1.2, 3.2) 1.6 (0.4, 2.7) 1.5 (0.8, 2.6) 2.1 (1.3, 3.3) 2.0 (1.2, 3.4)Nn physician in charge 1.(1' I.O' I.O' i .0 ' 1.0'" LO '

'Cnnfu'^inn, depress inn, or emotioFial distress.'^'Cnntrols lor co\ ari a tes with independent associaiii>n with respective symptom (at /' - (1.05) as shown in 'I a hie 3.' indica tes reference category.Note: Unadiusted and adiusted models est imated hy logi>tii' regi'ession. .All regressions control f o r s u r \ e \ ' sampl ing

strata.

and statistically insignificant differences were ob-served for management of dyspnea.

Respondents vvhii reported an early expecta-tion of death and those who reported lower priorpatient health status were also more likely to re-port adequate physician responsiveness to symp-toms. The effect of continuity in the physician-pa-tient relaliotiship was less evident, Dittereiices inrepoi'ted symptom responsiveness betweenphysicians without a relationship to the patientprior to the hospital stay and those with a priorrelationship were small and generally not statis-tically significant, except in the case oi pain con-trol.

Our findings suggest that it may be importantfor hospitals to organize services to ensure thateach patient is assigned one or more physiciansti.) co<irdiiiate cai'e, but ensuring that patients'commtmity-based physicians play that role maynot be essential. Tine finding that expectation ofdeath prior (o admission was associated withhighor ratings of physician resp(.>nsiveness un-derscores the importance of early and effectivecfimmunication between care providers atid fam-ily members about the patient's illness and like-lihood of survival. We dt) not have data to ad-clress whether family or provider expectationsabout sLu-vivai led to greater emphasis on pallia-tive cave, althoLigh this is a clear possibility.

It is noteworthy that several factors were notgenerally associated with reports of physician re-sponsiveness to symptoms. Specifically, we didnot find that race/ethnicity or indicators of so-cioeconomic statLLs or insurance coverage wereassociated with responsiveness. These findingsare reassuring and are not consistent with prior

findings of iowei' quality oi care and pain controlfor poor and minority patients.^''•^" We did, how-evej", find differences in physician symptom re-sponsiveness among the study hospitals in mul-tivariate analysis contixilling ft)r patient andnexf-of-kin characteristics.

Our nnalvsis has se\'eral limitatiotis. First, wereported on a retrospective survey ol SLU'\'i\'ingnext ol kin, which rec|uireci recall (on a\'erage ap-proximately 6 months) and reflected perceptionsof individuals without medical traininj^. Prior ev-idence suggests that compared to patients, sur-rogate respondents may report more symptomsand be more critical of caregivers, but we haveno reason tt) belie\ e that this e\ ide-nce sLiggcsts abias in the association between reported physi-cian-patient relationship and reported adeqLiacyof symptom management. Nevertheless, we at-tempted to control for differences iu the subjec-tive judgments among respondents by measuringI'espondent characteristics that might be corre-lated with perceptions (for example, ago, gender,and educational attainment). We also limited ouranalysis to cases whore interviewers rated re-sponding next of kin as knowledgeable. Whiiefurther work in this area would bo strongthonedby incorporating objoctivo moasuros, we noto thatfamily mombors are uniquely positioned to ob-ser\e the unfolding course of caro and thoir ex-periences aro important in thoir own right.

Socond, tho study is cross sectional and causalinferences must be macie with caution. Whilo wohavo suggested that having a physician in chargolod to more satisfactory symptom management,it may ha\'o been that better symptom control lodto tho perception that a physician was in chargo.

540 CANTOR ET AL.

The issue of causality can be addressed more of-fectively otily throLigh longitudinal or interven-tion research. Third, we did not describe the or-ganization of care beyond whether a physicianwas reported to be in charge. The role or re-sponsivonoss of nurses or othor caregi\'ors whomight have been involved in the symptt)m mau-agoment, and tho characteristics of tho physiciansproviding caro such as specialty or lovel of ti'ain-ing. In particular, the growing use of hospitalistsis worthy of examination in furthei' research. Fi-nally, our sLtrvey was limited to five teaching hos-pitals in oue region. Experiences in other settingsmight differ from our study facilities, and withintheso hospitals wo cann(.>t rulo out the possibilitythat tionrespondents might have reported differ-ently than respondonts.

In conclusion, symptom control and physicianrosponsivonoss to symptoms aro serious con-cerns for hospitalized patients near the end oflife and tboir families. Our findings suggest thatonsLU'ing that ouo or more physicians are clearlyin charge of care of oach pationt may bo an im-portant step toward improving the quality ofcare ft)r hospitalized dying pationts. Future re-search on palliative caro interventions shouldexamine the potential contribution of a primary,coordinating physician. In the absence of con-trary evidence, our study suggests that it is ad-visable for hospital-based palliative care inter-

ventions to encouragecoctrdinating physician.

a stron<i role for a

ACKNOWLEDGMENTS

The authors wish to thank Dr. Joan Teno whoassisted in the development of tho study ques-tionnaire; Eve Weiss and Jacki Smith who playedvital roles in the implementation of this study;Mtinifa English who provided programming as-sistance, and Professor Donald Hoover who pro-vided statistical advice. The Fan Fox and LeslieK. Samuels Foimdation and the United HospitalFLind provided financial support for this study.Dr. Blustein was supported by a Steven Char-noy Vladeck Junior Faculty Fellowship at NewYork University, and Dr. Carlson was supportedin part by National Institutes of Mental Healthtraining program nLuiibor MFII6242 wheu heserved as a postdoctoral follow atUniversitv.

Rutgers

REFERENCES

1. SUPPORT Principal Inxestigatois: A cnntrnlk'd trialtn imprnve care lor serinusly ill hospitalized patients:the Study tn Understand Prognoses and Preferencesfnr Outcomes and Risks of Treatments (SUPPOf-!T].JAMA 1993;274:l.59! -1598.

2. Bensnii JM, Cantor JC, Lynn J, Teno J: A national per-spective on dying in America: Does place of deathmatter? Brief Report. J Am Geriatr Soc May 2000;4S:523()-5231.

3. Hearn J, lligginsnn IJ: Dn specialist palliative careleams improv'c outcomes ior cancer patients? PalliatMed 1998; 12:317-332.

4. Pritchard RS, Fisher ES, Teno |M, Sliaip JM, RedingDJ, Kraus WA, Wennberg JE, l..ynn j : Influence oi pa-tient preferences and local health system characteris-tics on the place of death, j Am Cieriatr Soc l'-)98;46:1242-12.^0.

5. Conmbs RH, May DS, Small GW: Inside Doctnring:Stages and Outcnmes in the Professional De\'elop~ment ol Physicians. New Ynrk: Praeger Publishing,19Sh.

h- Degner LV. Beatnn JT: Life-Death Decisions in fiealthCare. Washingtnn. D.G.: I iemisphcrc Publishing,1987.

7. Trask At^, Faber DR: The intensix e care unit^—Who'sin charge^ Arch Surg IWtl;l25:l 105-1 !0S.

8. Bel/berg H: i he iTitensi\'e care unit: Who's in charge?Clin Neurosurg 198'-';35:63-67.

••). Manian FA: Whiter c(»nlinuity of care? N Eiigl J MedP>W;340:1362-1363.

10. 1 ligginson 1, Priest P, McCarthy M: .Are tx'renved fam-ily memt^els a \aiid proxy lor a patient's assesSTiientof dying? Soc Sci Med 1994;3S:553-5S7.

11. Desbiens N/\, Mueller-Ri/ner N: How well do surro-gates assess the pain oi" sei iniisly iH patients? Crit CareMed 2()00;28:1.347 1352.

12. Lobchuk MM, Kristianson L, LVgncr 1.,, BInod P,SlndnJA: Perceptinns nf symptom distress in king cancel-patients: 1. Congruence betvAeen patients aiul familycaregivers. J Pain Symptom Manage 1997;i4:i36-l46.

13. Zuckerman C, .Vlackinnon A: The challenges oi car-ing fnr patients neai- the end of life: Findings Irnm thelinspital Palliative Care lnitiati\e. New Ynrk: Uniti-dHospital Fund, I'-NS.

14. Linder Pel/ S: Social psychological determinants ofpatient satisfaction: A test nl live hypntheses, Snc SciMed 1982;lh:583-589.

\5. Williams B, Cnyle J, 1 !eal\ D: The meaning of palienisatisfactinn: An explanatinn of high reported levels.Snc Sci Med t998;47:13.5|-1359.

16. 1 lays Ri:), Shaul JA, Williams VSL, Lubalin JS, Hai-ris-Knjetin LD, Sweeny SI', Clearly I'D: Psychnmetricprnperties nf the CAFIPS'FM L(l survev measures.Med Care 19^)Q;37:MS22-MS3L

17. Da\is K, Cnllitis KS, Schnen C, Morri-- C: Choice Mat-ters: Lnrollee^' \ ievvs nf their health plans. Health Aff(MilKvnod) 1943;I4:99-I 12.

PHYSICIAN RESPONSE TO SYMPTOMS IN DYING PATIENTS 541

IS. Gaw.inde AA, Blendon RJ, Brodie M, t"icnson JM,Levit L, liugick L: Docs dissatisfaction with healthpl.ins stem frnni ha\ ing no choices? I lealth Aff (Mill-wond] 19^7; 17:184-194.

19. Kahn KL, Pearsoii Mi., Flarrison FR, Desmnnd KA,Rogers Wll, Rubenstcin LV, Brook KH, Kcelcr I'B:i lealth idiv for black and poor hospitalized Medicarepatients. |AMA 1994:271:1169-1174.

20. Tndd KI 1, Samaron N, hinffman JR: Lthnicity as a risklactnr lor inadequate I'lnergency di'parlment analge-sia. lAMA |W3;269;|537-I539. '

Address repi"int requests to:/()(•/ Cantor, Se.D.

Processor and DirectorRutgers Center for Staie Heallh Policy

Ml George StreetSuilc 400

NcTV Brunswick, Nl 08901

L-mait: jcautor@cshp.rutgors.edu