Professional Documents
Culture Documents
Factor
Factor
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at .
http://www.jstor.org/action/showPublisher?publisherCode=springer. .
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
Springer is collaborating with JSTOR to digitize, preserve and extend access to Quality of Life Research.
http://www.jstor.org
h Quality of Life Research 12: 167-176,2003. 167
l t 2003 Kluwer Academic Publishers. Printed in the Netherlands.
Abstract
Table1. Sample characteristicsof the close family members utilizes a five-point Likert scale ranging from 1,
(respondents)and the patients, absolute frequencies and very satisfied;2, satisfied;3, undecided;4, dissa-
percentages (n = 181)*
tisfied;and 5, very dissatisfied.
Respondents Patients
Statistical analysis
Age median(range) 56.4 (22-83) 68 (42-93)
N % N % The questionnairesof the 181 respondentscon-
Gender tained only a total of 39 missingvalues for all 20
Male 57 31.7 94 52.2 items. Most frequently,questionnairescontained
Female 123 68.3 86 47.8 only one or two missing values. Only two items
Relationto patient had more than 10 missing values:item 5 with 14
Spouse 114 63.7 - - (8%) and item 15 with 12 (7%). A missingvalue
Parent 1 0.6 - -
Child 58 32.0 - -
analysisby means of the missingvalue module in
Sibling 3 1.7 - - SPSS 10.0 showed a non-significantvalue of
Friend 3 1.7 - - Little's MCAR test (X2= 431, df= 458, p =
Childrenlivinghome 0.008),i.e. this indicatesthat the missingvaluesare
Yes 58 32.4 - -
completelyrandom.The missingvalueswere then
No 121 67.6 - -
Education estimatedusing the EM-algorithm.
,<7 32 17.9 64 35.6 To examine the dimensionalityof the FAM-
8-10 64 35.8 65 36.1 CARE items we performedboth a factor analysis
11-12 42 23.5 21 11.7 and an analysis using Mokken's non-parametric
< 13 41 22.9 30 16.7 IRT model.Oursamplesize of 181is just adequate
Work
Does not work 90 50.3 - -
to performa factoranalysis.As a ruleof thumb,it
Part time 20 11.2 - is recommendedthat the sample size is about 10
Full time 69 38.5 - - times the numberof items, i.e. 200 for 20 items.
Cancerorigin The computerprogramSPSSwas used to perform
Gastrointestine - - 84 46.7 a factor analysis. The extraction method was
Lung - - 25 13.9
Breast/female genitals - - 20 11.1
principalcomponent factoring with Kaiser'scri-
Prostate - - 16 8.9 terion. The initial solution was rotated (varimax
Others - - 35 19.4 with Kaiser normalization).The correlationbe-
Placeof death tween the factors and the items are indicatedby
At hospital - - 118 65.6
- -
the factor loadings, and the communalities(h2)
In nursing home 22 12.2
At home - - 40 22.2
show the proportionof the variancein each item
that is explainedby the factors.
* n = 181, i.e. all respondentswith valid values for the
A reliabilityanalysis of the FAMCARE Scale
FAMCAREScale.At most two respondentsor patientshave was performedusing the scale proceduresin SPSS
missingvaluesfor the demographicvariablesin this table. 10.0.
To furtherexaminethe psychometricproperties
third (32%) were childrenof the cancer victims. of the FAMCARE Scale, Mokken's non-para-
Only seven of the respondentswereparents(1%), metric latent trait model for one-dimensional
siblings(2%)or friends(2%) of the cancervictims. scaling was used [43, 44]. Central in the item
analysis is Loevinger'sH-coefficientof homoge-
The questionnaire neity or scalability.For each pair of items, it is
defined as Hgh = 1 - (Fgh/Egh), where Fgh is the
A questionnairewas mailedto the home addressof sum of observederrorsaccordingto the Guttman
each respondent.It includedquestionsabout how scale model, and Egh is the expected numberof
satisfiedclose family memberswere with the ad- errors from the statistical independencemodel.
vanced cancercare provided,as measuredby the The scalabilityof a single item with respectto the
FAMCARE Scale [1, 17, 38] 1 month after the other items in the scale is definedby Hg and the
time of death. The 20-item FAMCARE Scale scalability of the total scale is measuredby H.
170
Molenaar et al. [44] introducedthe weighted H- also has to be at least 0.3. Then the pair is ex-
coefficient,Hwgt,the defaultin the currentversion pandedwith the item that maximizesthe H-value
of the Mokken scalingprogram(MSP), the com- for a scale of three items. This process continues
puter programused to evaluate the FAMCARE until all items with Hj > 0.3 are included.
Scale. High H-coefficientsindicate good chances The FAMCARE items are scored so that high
that the assumptionsof a Mokken scale are satis- values indicate low satisfaction with the care
fied.A set of itemsconstitutesa scale if all Hghare component(see footnote a in Table 2). To make
greaterthan zero and if every item coefficientof the FAMCARE Scale easier to interpret, we
scalability,Hg, is largerthan a constantc, set to at wanted high scores to indicatehigh levels of sat-
least 0.30. All Hg and the H should be significan- isfaction, and the scale scores to vary between0
tly greaterthan zero accordingto a given level of and 100. The following SPSS command creates
significance. The total scale should have an our 19 item FAMCARE Scale (item 14 was ex-
H-value of at least 0.30 to form a weak scale. cludeddue to reasonsreportedin Table 5):
H-values between 0.40 and 0.50 indicate average
compute famcare
scales,and valuesabove0.50 indicatestrongscales. = 100- (((sum(item I to item 13, item 15 to item 20))
The MSP may be used to test predefinedscales - 19)/64) x 100).
or search for scales in a set of items. The search
option is a bottom-up procedure and starts by First, the sum of the 19 items were computed.
finding the two most homogeneous items, i.e. From this, the minimumobservedsum of 19 was
which give the highestH-statisticwith H = 0.3 as deductedand the resultdividedby 64, the rangeof
the minimumacceptable.Each item statistic, Hj, the observed sum scores (83-19). To make the
Mean Std.b 1 +2 1 4+ 5 5
1 The patient's pain relief 2.06 1.03 74.6 33.2 9.3 3.9
2 Information provided about the patient's prognosis 2.51 1.22 55.8 23.2 23.2 7.7
3 Answers from health professionals 2.11 1.02 68.5 32.6 9.9 2.2
4 Information given about side effects 2.72 1.23 47.5 16.6 23.8 12.7
5 Referrals to specialists 2.45 1.16 61.9 18.8 16.6 8.8
6 Availability of a hospital bed 1.91 1.19 81.8 48.1 13.8 6.6
7 Family conferences held to discuss the patient's illness 2.55 1.29 56.9 22.7 22.7 12.7
8 Speed with which symptoms are treated 2.47 1.22 62.4 21.0 20.0 10.5
9 Doctor's attention to patient's description of symptoms 2.05 1.06 73.5 35.4 9.4 4.4
10 The way tests and treatments are performed 2.07 1.00 72.9 32.6 9.9 2.2
11 Availability of doctors to the family 2.31 1.28 65.2 32.0 19.3 9.4
12 Availability of nurses to the family 1.68 0.86 84.5 51.9 2.7 1.7
13 Coordination of care 1.99 1.03 74.6 38.1 9.3 2.8
14 Time required to make a diagnosis 3.09 1.34 39.8 11.6 38.1 22.7
15 The way the family is included in treatments and care decisions 2.30 1.12 66.3 24.3 14.9 6.1
16 Information given about how to manage the patient's pain 2.48 1.22 59.7 22.1 20.0 10.0
17 Information given about the patient's tests 2.51 1.25 59.1 22.1 22.1 10.5
18 How thoroughly the doctor assesses the patient's symptoms 2.39 1.09 59.7 22.1 17.1 3.9
19 The way tests and treatments are followed up by the doctor 2.24 1.06 66.3 26.0 12.2 4.4
20 Availability of the doctor to the patient 2.15 1.07 68.5 31.5 11.6 3.9
a The response categories are: 1. Very satisfied, 2. Satisfied, 3. Undecided, 4. Dissatisfied, or 5. Very-dissatisfied.
b
Std.: standard deviation.
c 1 + 2: Includes the two highest response categories: very satisfied and satisfied, while 1 includes the
highest response category: very
satisfied.
d 4 + 5: Includes the two lowest response categories: very dissatisfied and dissatisfied, while 5 include the lowest category: very
dissatisfied.
171
scale vary from 0 to 100, the resultwas multiplied Removing item 14 reducesthe eigenvalueof the
by 100. The final operation, deductingthe scale weakestfactor from 1.22 to 1.08.
score from 100, reversesthe scale, so that high Since a factor analysis does not estimate the
values mean high satisfactionwith care. numberof factors, the conclusionon the dimen-
sionality will always be somewhat subjective.In
this case the strong first factor may be taken as
Results an indicationof one-dimensionality,but the two
factorswith eigenvaluesabove 1.0 may give some
Satisfaction with care based on single items reasonsfor doubt (Table 3).
To further explore the possibility of multidi-
We startedby examiningthe results for separate mensionalitywecomparedthefactorsolutionin our
items.Table 2 gives an overviewof the 20 itemsin study with Kristjanson's[1] conceptualstructure,
the FAMCARE Scale and descriptivestatistics. and her somewhatdifferentempiricalstructure.In
Low item values indicate high satisfaction with Table 4 the itemsare classifiedaccordingto Krist-
care.Most of the respondentsreportedto be either janson's [1] four-dimensionalconceptualstructure
'very satisfied'(value 1) or 'satisfied'(value 2) on and the threefactorsin our Norwegiandata.
most of the 20 items. The following items: Our first factor may be labelled 'Generalcare'
'Availabilityof nurses to the family' (item 12), and crosscuts all four conceptual dimensionsin
'Availabilityof a hospital bed' (item 6), 'Coordi- Kristjanson's[1] study. Four of her information
nation of care' (item 13), 'Doctor's attention to items loaded stronglyon our secondfactor, which
patient's descriptionof symptoms'(item 9), and we have labelled'Information',but it also picked
'The patient's pain relief' (item 1) indicated the up strongloadingsfrom items in her psychosocial
highest satisfaction with care (i.e. mean scores and physical care dimensions.Our third factor,
from 1.67 to 2.06). As many as half of the re- 'Physical care', picked up three items of her
spondentsreportedto be 'very satisfied'with the physicalcare dimension,but also one item on in-
'Availability of nurses to the family' and the formation giving and one on care availability
'Availabilityof a hospitalbed'. On the otherhand, loaded stronglyon this factor.
the items: 'Time requiredto make a diagnosis' The most reasonableinterpretationof the factor
(item 14), 'Informationgiven about side effects' analysisis that the FAMCAREitems tap a single
(item4), 'Informationprovidedabout the patient's underlying dimension. This is indicated in the
prognosis'(item 2), and "Informationgiven about large differencein eigenvaluesbetween the first
the patient'stests" (item 17) indicatedthe lowest and the two remainingfactors in the initial solu-
satisfactionwith care (i.e. mean scores from 3.09 tion. If the two weakfactorsactuallyrepresenttwo
to 2.51). However,only a few of the respondents additionaldimensions,the factor structurein the
(7.7-12.7%) reportedthat they were 'very dissat- Norwegiandata does not easily conformto either
isfied'with respectto these items, except for item Kristjanson's [1] conceptual structure for the
14 with 22.7% 'verydissatisfied'. FAMCARE items or to her somewhat different
empiricalstructure.
Dimensionality - factor analysis
Using the MSP to searchfor scales
To explore the dimensionality of the 20
FAMCARE items we performeda factor analysis The MSP may be used to test predefinedscales or
and comparedour resultsto those from Kristjan- searchfor scales in a set of items. Becauseof the
son's [1] study. The factor analysisin the Norwe- lack of consistencybetween conceptualstructure
gian sampleresultedin one very strongfactorwith and the factor analysis, the search option was
an eigenvalueof 10.95,and two weak factorswith used. The results from the MSP analysis are
eigenvalues of 1.52 and 1.22. If item 14 were reportedin Table 5 along with the resultsfrom a
eliminated, the remainingitems would all have reliabilityanalysisof the scale. The resultingscale
their highestfactor loadings on the first factor in consistsof all itemsbut item 14:'Timerequiredto
the initial factor solution, i.e. before rotation. make a diagnosis',that was excludeddue to a too
172
Table 3. A factor analysis of the FAMCARE items with factor loadings and communalities, n = 181*
* Extraction method: principal component factoring with Kaiser's criterion, rotation: Varimax with Kaiser normalization. High factor
loadings (>0.60) are given bold.
Fl": factor loadings for the first factor in the initial (unrotated) solution.
Fl, F2, and F3: factor loadings for rotated solution.
h2: communalities, the proportion of the variance in each item which is explained by the factors.
Table 4. Cross-classification of the conceptual and empirical structures of the FAMCARE items
low coefficient of homogeneity (Hwgt= 0.27). The The FAMCARE Scale was initially computed as
19-item scale is a strong one both in terms of the sum of the scores on the 19 items, rescaled to
scalability (Hwgt= 0.59), and the reliability of the vary between 0 and 100 and reversed so that low
scale (Cronbach'sa = 0.96). scores on the FAMCARE Scale indicate low levels
173
Table5. Scalabilityand reliabilityanalysisof the FAMCAREitems, n - 181. The itemsare orderedby sampledifficulties(mean
values)
H,,,,gt Corr. R2
10
tivelyskewed(skewness= -0.41, standarderror= C
0
0.18;kurtosis= -0.73, standarderror= 0.36).The 20%4)
shape of the distributionis betterseen in the his-
0
togramof the FAMCARE Scalein Figure 1. C ~ ~ ~
210%_
Discussion
0%
The high level of satisfactionwith care for dying
cancer patients reported by family members or
0 20 40 60 80 100
close friends in our study is in accordancewith
most studies of satisfaction with palliative care FAMCARE Scale
[1, 9, 18-20, 29, 32, 39]. In terms of single items, Figure 1. Distribution of the 19 item FAMCARE Scale (his-
whichmay be of clinicalinterest,we foundthat the togram). Descriptive statistics for the FAMCARE Scale: mean:
respondentsreportedhighest level of satisfaction 62.53; standard deviation: 25.24; scale range: 0-100.
174
with the followingitems:'Availabilityof nursesto of high satisfactionthat we have seen our study
the family', 'Availability of a hospital bed', shares with many others may also be due to the
'Coordinationof care', 'Doctor's attentionto pa- reluctanceof close family members or informal
tient's description of symptoms', and 'The pa- caregiversto express negative perceptionsabout
tient'spain relief. On these itemsas muchas three care, because of their belief that nothing would
out of four reportedto be either'verysatisfied'or change as a result of complainingabout the ser-
'satisfied'. vices. To enhance their self-esteem and also to
The numberof respondentsdissatisfiedwith the reducecognitivedissonance,family membersmay
carecomponentswas low for most of the items.In also have retroactivelyinflatedtheir evaluationof
our study, only 3.9% of the respondentsreported the servicesthat reflecton the wisdomof theirown
to be 'very dissatisfied'with 'The patient's pain behaviouror judgementin choosing these services
relief, whichmightbe seen in rathergreatcontrast [20]. Alternatively,high satisfactionmight be an
to anotherfinding[33]in whichhalf of the sample accuratereflectionof the qualityof the healthcare
of familymembersto cancervictimsreportedthat they receive.
the patient experienced'very distressing'pain at The high percentagesof 'satisfied' and 'very
some time in the last year of life. Furthermore,we satisfied'in our studyindicatepoor discrimination
found that the items on informationabout side at high levelsof satisfaction.The responsescale of
effects,the patient'sprognosis,pain management, 'very satisfied'to 'very dissatisfied'has also been
the patient'stests, and the item on familyconfer- criticized in an earlier work [36]. Patients and
ences to discussthe patient'sillness, indicatedthe family membershave reporteddifferentsatisfac-
lowest satisfactionwith care, with about one of tion with medical care based on whetherthe re-
four reporting'dissatisfied'or 'very dissatisfied'. sponse scale was 'poor to excellent' or 'very
Complaintsabout poor informationand commu- dissatisfied'to 'very satisfied',and Ware [37] re-
nication were also reportedin a national popula- commendedthe use of the formerversion.
tion-basedinvestigationincludinga large sample A third methodologicaldifficultydeals with a
of family membersof dying cancer patients [33], retrospectivevs. a prospectiveapproach in col-
whereas many as half of the respondentsdid not lectinginformationabout family satisfactionwith
get all the informationthey wantedabout the pa- care for the dying. Interpretationof the results
tient'smedicalcondition. might be difficultdue to differencesbetween the
Our findingof a strong one-dimensionalFAM- time period assessedprospectively,i.e. before the
CARE Scale (Hwgt= 0.59, Cronbach's ot= 0.96) death of the deceased [1, 9, 17-20, 38, 39] and
supports the conclusion in Kristjansonand her those assessedretrospectively,i.e. after the death
colleagues'study[17].The largea is due to the high of the deceased [20, 29, 32, 33, 35, 46]. Small
numberof items and the strong inter-itemcorre- samplesizes may also impactthe results[1, 17, 32,
lations. Thus, the MSP analysissupportsthe one- 35, 38, 39]. Retrospectivestudies have collected
dimensionalinterpretationof the factor analysis. data at quitedifferentpointsin time afterthe loved
However, since their factor analysis results were ones' death. One seriousproblemwith the retro-
only sparselydescribedin the text, a moredetailed spectivedesign in satisfactionresearchis that the
comparisonof our findingwiththeirsis impossible. problemswith recall and the subjectivitybias in-
In our study, item 14: 'Time requiredto make a crease along with the amount of time that has
diagnosis',had to be excludedfromthe scaledue to elapsedfrom the death until the collection of the
the results from both the factor analysis and the data. On the other hand, the retrospectivedesign
MSP analysis. makes it possible to have a representativesample
There are several methodologicalproblems in of all people who died, not just those who were
studyingsatisfactionwith care.One problemin the identifiedto be terminallyill or receivingpalliative
evaluation is social desirability[45], i.e. the re- care services[33, 46]. When satisfactionis exam-
spondentsare answeringin a way they think the ined beforedeath the designis also retrospectiveif
society expect them to answer, and they may be there is a time period between the hospital-care
concernedabout how the care may changeshould service received and the point in time for data
they respondwith anythingbut praise.The finding collection. Further, in prospective studies the
175
15. Ringdal GI, Jordh0yMS, Ringdal K, Kaasa S. Factors spouse on effectsof the serviceon the patient. Postgrad
affectinggrief reactionsin close familymembersto indi- Med.J 1979; 55: 517-522.
vidualswho havedied of cancer.J Pain SymptomManage 35. ParkesCM. Terminalcare:Evaluationof in-patientservice
2001;22: 1016-1026. at St Christopher'shospice. Part II. Self-assessmentsof
16. Linder-PelzS. Toward a theory of patient satisfaction. effectsof the serviceon survivingspouses.PostgradMed J
Soc Sci Med 1982;16:577-582. 1979; 55: 523-527.
17. KristjansonLJ, SloanJA, DudgeonD, AdaskinE. Family 36. Ware JE. Effectsof acquiescentresponseset on patient
members'perceptionsof palliativecancercare:Predictors satisfactionratings.Med care 1978;16:327-336.
of family functioningand family member'shealth. J Pal- 37. WareJE. Methodsfor measuringpatientsatisfactionwith
liativeCare 1996;12: 10-20. specificmedicalencounters.Med Care 1988;26: 393-402.
18. HigginsonI, WadeA, McCarthyM. Palliativecare:Views 38. KristjansonL, Leis A, Koop P, CarriereKC, MuellerB.
of patientsand theirfamilies.Br MedJ 1990;301:277-281. Familymembers'care expectations,care perceptions,and
19. HigginsonI, PriestP, McCarthyM. Are bereavedfamily satisfactionwith advancedcancercare:Resultsof a multi-
membersa validproxyfor a patient'sassessmentof dying? site pilot study.J PalliativeCare 1997;13:5-13.
Soc Sci Med 1994;38: 553-557. 39. JarvisH, BurgeFl, Scott CA. Evaluatinga palliativecare
20. FakhouryW, McCarthyM, Addington-HallJ. Determi- program:Methodologyand limitations.J PalliativeCare
nants of informalcaregivers'satisfactionwith servicesfor 1996;12:23-33.
dyingcancerpatients.Soc Sci Med 1996;42(5):721-731. 40. OsseBHP,Vernooij-Dassen MJFJ,de VreeBPW,SchadeE,
21. BarzelaiLP. Evaluationof a home based hospice.J Fam Grol RPTM.Assessmentof the need for palliativecareas
Practice1981;12:241-245. perceivedby individualcancerpatientsandtheirfamilies.A
22. CreekLV. A homecarehospiceprofile:Description,eval- reviewof instrumentsfor improvingpatientperceptionin
uation, and cost analysis.J Fam Practice 1982; 14: 53- palliativecare.Cancer2000;88: 900-911.
58. 41. Jordh0y MS, Fayers P, Saltnes T, Ahlner-ElmquistM,
23. LongmanAJ, Atwood JR, ShermanJB, BebedictJ, Shang Jannert M, Kaasa SA. Palliative-careinterventionand
TC. Care needs of home-basedcancerpatientsand their death at home: A clusterrandomisedtrial. Lancet 2000;
caregivers.Quantitativefindings.CancerNurs 1992; 15: 356:888-893.
182-190. 42. Jordh0yMS, KaasaS, FayersP, OvrenessT, UnderlandG,
24. KristjansonLJ, Atwood J, Degner LF. Validityand reli- Ahlner-Elmquist M. Challengesin palliativecare research;
abilityof the familyinventoryof needs(FIN): Measuring recruitement,attritionand compliance;experiencefrom a
the care needs of families of advancedcancer patients. randomisedcontrolledtrial.PalliativeMed 1999;13:299-
J Nurs Measurement1995;3: 109-126. 310.
25. McCuskerJ. Developmentof scalesto measuresatisfaction 43. SijtsmaK, DebetsP, MolenaarIW. Mokkenscale analysis
and preferencesregardinglong-termand terminalcare. for polychotomousitems:Theory,a computerprogramand
Med Care 1984;22: 476-493. an empiricalapplication.QualQuan 1990;24: 173-188.
26. HilemanJW, LackeyNR, HassaneinRS. Identifyingthe 44. MolenaarIW, Debets P, SijtsmaK, Hemker BT. MSP
needs of home caregiversof patientswith cancer. Oncol User'sManual.A Programfor MokkenScaleAnalysisfor
Nurs Forum 1992;19:771-777. Polychotomous Items. Version 3.0. Groningen, The
27. SteeleRS, Fitch MI. Needsof familycaregiversof patients Netherlands:iec ProGAMMA,1994.
receivinghomehospicecarefor cancer.OncolNurs Forum 45. SudmanS, BradburnNM. Asking questions.San Fran-
1996;23: 823-828. cisco, CA: Jossey-Bass,1983.
28. HarringtonV, LackeyNR, Gates MF. Needs of caregivers 46. FakhouryWKH, McCarthyM. Can the experienceof
of clinicand hospicecancerpatients.CancerNurs 1996;19: caring at home affect carers'retrospectiveevaluationof
118-125. communitycareservices?ScandJ CaringSci 1998;12:179-
29. DawsonNJ. Need satisfactionin terminalcaresettings.Soc 185.
Sci Med 1991;32 (1): 83-87. 47. ParkesCM. The firstyear of bereavement.A longitudinal
30. Hampe SO. Needs of the grievingspouse in a hospital study of the reactionsof Londonwidows to the death of
setting.Nurs Res 1975;24: 113-120. theirhusband.Psychiatry1971;33: 444-467.
31. Houts PS, Yasko JM, HarveyHA, et al. Unmet needs of 48. FashingbauerTR, DevaultRA, Zisook S. Developmentof
personswith cancerin Pennsylvaniaduringthe period of the TexasInventoryof Grief.Am J Psychiatry1977;134(6):
terminalcare. Cancer1988;62: 627-634. 696-698.
32. Keizer MC, Kozak JF, Scott J. Primarycare providers'
perceptionsof care. J PalliativeCare 1992;8: 8-12.
33. Addington-HallJ, McCarthyM. Dying from cancer:Re- Addressfor correspondence:Gerd Inger Ringdal, Department of
sults of a nationalpopulation-basedinvestigation.Pallia- Psychology, Faculty of Social Sciences and Technology Man-
tive Med 1995;9: 295-305. agement, NTNU, 7034 Trondheim, Norway
34. ParkesCM. Terminalcare:Evaluationof in-patientservice Phone: +47-73-591960/597486; Fax: +47-73-591920
at St Christopher'shospice. Part I. Views of surviving E-mail: gerd.inger.ringdal(svt.ntnu.no