Fax: (011) 45-3531-2071
A longitudinal study of palliative care
Patient-evaluated outcome and impact of attrition
Article first published online: 8 MAR 2005
Copyright © 2005 American Cancer Society
Volume 103, Issue 8, pages 1747–1755, 15 April 2005
How to Cite
Strömgren, A. S., Sjogren, P., Goldschmidt, D., Petersen, M. A., Pedersen, L., Hoermann, L. and Groenvold, M. (2005), A longitudinal study of palliative care. Cancer, 103: 1747–1755. doi: 10.1002/cncr.20958
- Issue published online: 4 APR 2005
- Article first published online: 8 MAR 2005
- Manuscript Accepted: 14 DEC 2004
- Manuscript Revised: 5 NOV 2004
- Manuscript Received: 15 JUN 2004
- Danish Cancer Society. Grant Number: PP 02 020
- Copenhagen Hospital Corporation
- palliative care;
- advanced-stage cancer;
- self-assessment questionnaires;
- quality of life;
- treatment outcome
The current article evaluated the course of patient-assessed symptomatology in specialized palliative care and tested for bias due to patient attrition in measures of initial symptomatology and treatment outcome.
Over 2 years, 267 consecutive, eligible patients were referred to a department of palliative care. Upon arrival, 201 patients consented to participate in a questionnaire-based evaluation of quality of life (QOL). Of these, 175 patients participated, and 142, 119, and 95 participated in the study at 1, 2, and 3 weeks, respectively. Weekly, participants completed the self-assessment questionnaires European Organization for Research and Treatment of Cancer QLQ-C30, Edmonton Symptom Assessment System, Hospital Anxiety and Depression Scale, and Multidimensional Fatigue Inventory. Physicians used the Mini Mental State Examination to evaluate cognitive function. Changes from the initial symptom scores for each week were calculated. Initial scoring and change after 1 week were tested for association with completion level, i.e., whether the patient completed questions at 1, 2, 3, or 4 time points.
High initial symptom intensity and significant improvements over time were observed for pain, lack of appetite, nausea/vomiting, sleeplessness, constipation, and overall QOL/well-being. For some symptoms, initial scores were significantly higher in patients who dropped out, but the changes over the first week were not significantly different between completion levels for any symptom.
Improvement in symptom intensity was identified. Dropout was associated with higher initial symptomatology but not with poorer outcome of palliative treatment. Cancer 2005. © 2005 American Cancer Society.
To constantly improve treatment in palliative care, evaluation of the care delivered is essential. However, longitudinal studies in palliative care are few, probably because of methodologic and ethical difficulties. For example, recruitment for palliative care studies is difficult in a population of frail patients, and attrition due to deterioration or death is very likely.1–4 Few descriptions are available concerning the quantitative, patient-assessed evaluation of the outcome of palliative care outside clinical trials.5–7 The current study is such an evaluation, based on consecutive patients with advanced-stage cancer admitted to a department for specialized palliative care.
In the Department of palliative Medicine at Bispebjerg Hospital (Copenhagen, Denmark), a prospective research project was conducted, aimed at all patients referred to the department, to evaluate palliative care and to develop and validate methods for such evaluations. The initial symptom characteristics according to the European Organization for Research and Treatment of Cancer Quality of Life Questionnaire (EORTC QLQ-C30), the Edmonton Symptom Assessment System (ESAS), the Hospital Anxiety and Depression Scale (HADS), and the Mini Mental State Examination (MMSE) have been described previously in a group of patients with advanced-stage cancer referred for specialized palliative care.8 It was documented that these questionnaires cover the symptoms and problems reported in the medical records of these patients, and thus evaluate the symptoms for which they receive treatment.9 The characteristics, levels, and changes in pain over the first week have also been reported.10 Previous results have shown that patients who did not participate in the study were older, had lower Karnofsky performance scores (KPS), and had shorter survival periods than participants.8 It was concluded that nonparticipants had more advanced-stage disease and this would have more pronounced symptomatology. Thus, the initial symptomatology, as assessed by the participating patients, may well be underestimated. There is no information in the literature to indicate whether patients who do not participate in formalized evaluation studies benefit from specialized palliative care in the same way, as do participants. A major concern would be that nonparticipating patients and “partial participants” might benefit less from palliative care than complete compliers, and that longitudinal studies would therefore produce positively biased results. Hollen et al.11 suggested a method to control bias resulting from attrition due to death. According to this method, at the day of death, a score of zero (i.e., minimal quality of life [QoL]) would be assigned. However, in the current study, many patients would be assigned minimal QOL or maximal symptomatology even in symptom areas where they were not disturbed before. Therefore, we did not use the Hollen et al. approach.
The aim of the current study was to evaluate the course of patient-reported symptomatology after referral to specialized palliative care. Furthermore, the problem of attrition is investigated and a simple model is presented and tested to compare information from patients at different levels of compliance.
MATERIALS AND METHODS
According to the referral criteria, patients were referred to the specialist palliative care unit for symptom control and/or end-of-life care planning by other hospital departments or by their physicians, had advanced-stage cancer for which no curative treatment could be offered, and had “pronounced palliative needs.” Patients could receive inpatient and outpatient treatment as well as palliative home care. Patients referred between June 1998 and June 2000 were included in the study, provided they were Danish speaking, ≥ 18 years old, and cognitively able and willing to give informed consent. The ethics committee approved the study. Palliative care involved physicians (in the areas of oncology, anesthesiology, internal medicine), nurses, social workers, chaplains, psychologists, physical therapists, and dieticians.
This is a prospective, longitudinal evaluation of consecutive patients admitted to specialized palliative care, using patient-completed, validated instruments. Clinical and sociodemographic characteristics were registered for all referred patients, and the physician assessed patients' KPS value12 once a week. The length of survival from the first encounter with the department was registered. On the day of first contact, the patients were approached for possible participation in the study. Consenting patients were given the self-assessment questionnaires ESAS,13 the EORTC QLQ-C30 (version 3.0),14 the HADS,15 the Multidimensional Fatigue Inventory (MFI-20),16 and were assessed by the physician using the MMSE.17 A participant was defined as a consenting patient who completed ≥ 1 item in the ESAS or ≥ 1 scale or single item of the EORTC QLQ-C30 (the 2 general questionnaires) at the day of first contact or the next day. This definition was used to motivate patients to participate even if they were willing or able to fill in only some of the items. A staff member could assist the patient in writing down the responses. The patients were assessed weekly for ≤ 13 weeks. On the last page of the questionnaire booklet containing the three latter questionnaires, the patients could state whether they had completed the questionnaires themselves or whether another person had assisted in writing down the responses. Furthermore, the patients were asked to identify any questions that they found difficult to understand, as well as questions which they considered to be objectionable. Completion of the questionnaires was to take place on the first or second day of contact with the department.
When a patient was admitted, the attending physician gave the inpatients the ESAS to complete immediately. This division of inpatient self-assessment in two parts was chosen from a belief that the combination of two different data collection strategies would enhance the probability of obtaining at least a partial description of each patient's symptomatology. The outpatients and patients in palliative home care, however, received the ESAS as a part of a questionnaire booklet that also contained the three other self-assessment questionnaires. This data collection strategy made subsequent weekly self-assessments possible, even if the patient did not visit the physician on the exact weekday.
The staff had no access to the completed questionnaires, to protect confidentiality and to avoid potential bias, which might arise if patients wished to communicate with the staff through their responses to the questionnaires. Patients were informed that the physician would not see the completed questionnaire and that if they wanted to discuss issues from the questionnaire with the staff, then they should contact the staff.
In the current study, assessments from the day of first contact or the next day (T0) and assessments made 1, 2, and 3 weeks later (T1, T2, and T3, respectively) are used. Changes in scores on the EORTC QLQ-C30, ESAS, HADS, MFI-20, and MMSE are reported as parallel outcome measures.
The ESAS and the EORTC QLQ-C30 are designed to cover physical as well as psychosocial aspects of the patient's health-related QOL. The version of ESAS published in 199113 was used. It consists of 9 visual analog scales, each comprising a symptom or a problem and each ranging from 0 mm to 100 mm, where 100 corresponds to maximal symptomatology. It was developed for patients in palliative care and has been validated for use also in nonhospice patients with cancer.18
The EORTC QLQ-C30 comprises 6 function scales, 3 symptom scales, and 6 single items that measure specific symptoms. The responses were converted to 0–100 scales according to the scoring manual.19 For single items, 0 corresponds to not at all, 33.3 to a little, 66.7 to quite a bit, and 100 to very much. For the function scales, high scores reflect better functioning, whereas high scores on symptom scales reflect high symptom levels. Thus, differences over time for function scales must have positive values to reflect a positive outcome for the patient, in contrast to the symptom scales and items.
The 14-item HADS measures anxiety and depression. Scores range from 0 to 21 points on each of the anxiety and depression subscales. According to the developers, 0–7 points on a subscale represent a noncase, 8–10 points represent a doubtful or possible case, and 11–21 points represent a case of anxiety or depression.15
The MMSE comprises 21 items that cover orientation, memory, and attention, as well as ability to name objects, follow verbal and written instructions, write a sentence spontaneously, and copy a figure.17 Patients can score 0–30 points, and scores ≥ 24 are generally considered to be normal.20
The MFI-20, developed for patients with cancer, contains 20 items in 5 scales representing general fatigue, physical fatigue, mental fatigue, reduced activity, and reduced motivation.16 With the permission of the developer, the original subscales ranging from 4 to 20 were transformed into subscales ranging from 0 to 100, where high scores indicate more fatigue.
In case of missing items within 1 of the 9 multiitem scales of the EORTC QLQ-C30 or the 5 scales of the MFI-20, a scale score was calculated if at least one-half of the items in the scale were completed.19, 21 For the scales of the HADS, there are no generally accepted guidelines for handling missing data within the questionnaire. A scale score was calculated provided that at least six of seven items in each subscale were completed. Reliability of the multiitem scales was assessed by Cronbach's coefficient α.22 Internal consistency estimates of a magnitude ≥ 0.70 were sought.23
The mean differences in symptom scores between T0 and T1, T0 and T2, and T0 and T3 (Δ1, Δ2, and Δ3, respectively) were calculated for patients who completed questionnaires at both time points.
Considering previous results concerning participation in the current study (that nonparticipants had poorer health),8 there might be a risk of obtaining positively biased estimates of treatment effect if patients who drop out have the least change in symptom intensity. Treatment response from T0 to T1 could be compared across subgroups of patients with different levels of compliance. To evaluate whether treatment response was related to the level of compliance, treatment responses from T0 to T1 were compared across subgroups of patients with different levels of compliance. The hypothesis tested was that if patients who drop out have less change in symptom intensity compared with patients who stay in the study, the best treatment response during the first week would occur in patients staying in the study through all 3 weeks, and the poorest treatment response would occur in patients dropping out between T1 and T2. The pattern of completion was registered, i.e., the 8 different ways in which participants could complete ≤ 4 questionnaires (e.g., T0 and T1; T0 and T2; T0 and T1 and T2). Initial scores could be compared regardless of completion pattern. Clearly, longitudinal self-assessment data were unavailable from patients who only completed the first set of questionnaires. Therefore, when examining change during the first week (Δ1), only patients participating more than once could be compared. Comparing the influence of compliance on initial symptom levels and change over time, the Δ1 scores for patients with uninterrupted completion from T0 to T1, T2, or T3 (e.g., T0 and T1, T0 and T1 and T2, or T0 and T1 and T2 and T3) were used, thus excluding data from the 15 patients who had interrupted patterns of completion (e.g., T0 and T2 and T3) (see Table 2). This ensured that the patients in the groups compared were the same at a given time point. The test method was inspired by the graphical analysis method presented by Hopwood et al.24 Data were analyzed using the SAS statistical package 8.02 (SAS, Cary, NC). Wilcoxon rank tests were used to test the changes in scores over time and for comparison between completion levels.
Of 278 patients referred, 10 did not speak Danish and 1 patient was < 18 years old. Sixty-six (25% of 267 eligible patients) did not consent to participate: 5 declined and 61 were not approached about the study (staff decided that 29 were too cognitively impaired, 3 were too depressed, 21 were too physically ill, and for 8 patients, the reason was not stated). Of 201 patients consenting to participate, 8 patients failed to complete questionnaires after all. Eighteen patients only completed questionnaires at Weeks 1, 2, or 3, and these results could not be used in the outcome analysis. One hundred and seventy-five patients (66% of eligible patients) completed enough items of the EORTC QLQ-C30 or the ESAS at T0 to be considered participants. Of these, 34 were referred from oncology, 4 from hematology, 32 from a pain clinic, 35 from internal medicine, 43 from surgical departments, and 27 from physicians.
Compliance rates for different time points and instruments are shown in Table 1. Instruments presented to the patients by physicians (such as the ESAS and MMSE at T0) or placed first in the self-assessment booklet (EORTC QLQ-C30) had the best compliance. At 4–13 weeks, although still acceptable for the EORTC QLQ-C30 and ESAS, the completion rates were considered too low for the HADS, MFI-20, and MMSE for reliable results. Therefore, only the results from the first 3 weeks (T0, Δ1, Δ2, and Δ3) are reported. There was a special problem with the completion of the MMSE at T1–T3, as the outpatients and patients in palliative home care did not necessarily see the physician on the weekday of questionnaire completion. This means that the completion rate for this instrument was quite low (16–29% of participants) after T0 for outpatients and for patients in palliative home care.
|Week||Participants alive||Compliance: questionnaires completed (% of participants alive)|
|0||175||144 (82)||168 (96)||135 (77)||130 (74)||162 (93)|
|1||167||117 (70)||121 (72)||107 (64)||109 (65)||74 (44)|
|2||142||90 (63)||97 (68)||46 (32)||45 (32)||56 (39)|
|3||117||79 (67)||81 (69)||41 (35)||45 (35)||50 (43)|
|4||103||61 (59)||64 (62)||22 (21)||24 (23)||30 (29)|
|5||89||51 (57)||49 (55)||21 (24)||21 (24)||19 (21)|
|6||79||50 (63)||49 (62)||21 (27)||21 (27)||19 (24)|
|7||69||43 (62)||41 (59)||17 (25)||17 (25)||15 (22)|
|8||61||35 (57)||29 (42)||14 (20)||15 (22)||8 (13)|
|9||56||37 (66)||35 (63)||14 (25)||14 (25)||14 (25)|
|10||51||31 (62)||24 (47)||14 (27)||14 (27)||8 (16)|
|11||50||26 (52)||22 (44)||12 (24)||12 (24)||9 (18)|
|12||47||28 (60)||26 (55)||14 (30)||14 (30)||10 (21)|
|13||46||27 (59)||27 (59)||11 (24)||11 (24)||12 (24)|
Patients spent an average 28 minutes (range, 2–120 minutes) to complete the self-assessment questionnaires at T0. The time consumption decreased to averages of 24, 22, and 19 minutes at T1, T2, and T3, respectively. Fifty-eight percent of the patients completed the questionnaires themselves, whereas the rest received some degree of help.
Table 2 shows the 8 completion patterns. Seventy-six (43%) patients completed all 4 questionnaires, 31 (18%) patients completed 3, 35 (20%) patients completed 2, and 33 (19%) patients completed only the first questionnaire. Hence, 81% of the 175 participants completed ≥ 2 questionnaires for the outcome analysis.
|No. of patientsa||T0||T1||T2||T3||ΣNb|
Patient sociodemographic and clinical data are shown in Table 3.
|Characteristics||No. of patients (%)|
|Mean/median age (range)||62.8/63 (37–91 yrs)|
|Mean/median survival (range)||94.2/35 (3–1217 days)|
|Mean/median KPS (range)||46.4/40 (20–80)|
|Head and neck||8 (4.6%)|
|Gastrointestinal tract||36 (20.6%)|
|Respiratory system||46 (26.3%)|
|Genitourinary system||29 (16.6%)|
|Locally advanced||20 (11.4%)|
Levels of Symptoms/Problems and Changes over Time
In Table 4, initial scores and changes over 1, 2, and 3 weeks for the ESAS and the EORTC QLQ-C30 are shown. Significant improvement was seen for many symptoms after ≥ 1 week. No symptoms worsened significantly over time, although the mean KPS value declined (the mean initial KPS value was 46.5 at T0 and the mean Δ3 was −5.3, P = 0.0034). The decline in KPS value was less in patients who participated at a given time point compared with patients who did not (e.g., Δ1 = −1.9 for 121 patients completing at T1, and Δ1 = −11.5 for 26 patients not completing at T1, P = 0.005).
|Quality of lifec||135||31.7||0.81||91||5.8d||72||4.5||66||8.7e|
As previously reported, significant decreases in pain scores were found during the first week of palliative care.10 For fatigue, high initial scores were found on the EORTC QLQ-C30 (Table 4) as well as on the MFI-20. Only modest reductions in fatigue scores were found over time, which were statistically significant only for the MFI-20 subscales general fatigue and reduced activity. A nonsignificant increase in mental fatigue was found (data not shown).
Except for an improvement in EORTC QLQ-C30 global health status/QOL in Δ1 and Δ3, function scales were unchanged over time. Lack of appetite was significantly alleviated in Δ1 (EORTC QLQ-C30) and in Δ2 and Δ3 (EORTC QLQ-C30 and ESAS). According to the EORTC QLQ-C30, nausea/vomiting declined significantly after 2 and 3 weeks, but this was not found in the ESAS nausea item. Dyspnea did not change over time. Constipation decreased significantly over 2 and 3 weeks. Initial ESAS inactivity was high. No significant improvement over time could be detected. No significant changes in depression scores over time were found (HADS or ESAS). The HADS revealed a small but significant improvement in anxiety (Δ1 = −0.7, P = 0.03; data not shown), but ESAS anxiety was unchanged.
Cronbach's α coefficients for the EORTC QLQ-C30 were > 0.70 for all scales except for cognitive function and nausea/emesis (0.60 and 0.69, respectively) (Table 4).
Table 5 shows initial scores and Δ1 when the patients are divided according to level of completion. Only scales/items with significant differences between levels of completion are shown. Patients who dropped out had more severe initial symptoms. In contrast, only the KPS value revealed significant difference in Δ1 for different levels of completion, i.e., patients dropping out had a decreasing KPS value.
|KPS (physician assessment)|
|T0 and T1||29||43.4||28||−7.1c|
|T0, T1, and T2||20||43.0||19||1.6|
|T0, T1, T2, and T3||75||50.9||71||−0.7|
|T0 and T1||28||47.9||23||6.5|
|T0, T1, and T2||21||40.8||19||2.3|
|T0, T1, T2, and T3||71||30.4||69||−0.4|
|EORTC physical function|
|T0 and T1||22||22.1||17||−0.4|
|T0, T1, and T2||16||27.1||16||−3.3|
|T0, T1, T2, and T3||63||36.1||61||−1.6|
|EORTC cognitive function|
|T0 and T1||22||37.9||15||6.7|
|T0, T1, and T2||16||55.2||16||−7.3|
|T0, T1, T2, and T3||64||58.9||61||1.6|
|T0 and T1||22||86.4||16||−6.3|
|T0, T1, and T2||16||77.1||15||−3.0|
|T0, T1, T2, and T3||65||71.6||62||−1.1|
|T0 and T1||22||56.1||15||−8.9|
|T0, T1, and T2||15||40.0||15||8.9|
|T0, T1, T2, and T3||65||31.8||62||5.4|
Two graphic examples of outcome are shown in Figures 1 and 2. The mean scores for ESAS dyspnea (Fig. 1) and ESAS pain (Fig. 2) at T0–T3 are shown for the 4 levels of (uninterrupted) completion. As illustrated in Figure 1, the mean ESAS dyspnea scores at T0, T1, and T2 are different between completion levels (P < 0.05). However, the differences between T0 and T1, as well as between T1 and T2, are not convincingly deviant, indicating that these changes are not dependent of completion level. In Figure 2 (ESAS pain scores), the initial scores are at the same level, and the curves appear parallel. Thus, for pain, scores initially as well as later seem independent of completion level.
Is it realistic to perform longitudinal studies in palliative care? The problems and drawbacks are evident. Regarding stringency in data and results in a population for which the median survival is approximately 1 month and the performance of surviving patients is declining, looking ahead might well be looking for trouble. Simon and Wittes25 recommended that ≥ 85% of patients should be evaluable for reliable results in clinical trials, and that inevaluability rates > 20% would often reflect inappropriate patient selection. This is true in many clinical settings. In palliative care research, this is an ambitious goal, as many patients, although still alive, are frail and often physically as well as cognitively impaired. If inclusion criteria were to be applied with expected survival time > 2 months, the included sample would no longer be representative of the patient population. Apparently, no longitudinal study in palliative care yet has managed to meet these criteria.26–29
A positive effect of palliative care has been found in a few uncontrolled, longitudinal studies of reasonably consecutive patients. Chiu et al.26 described the change in staff-assessed severity of symptoms in 232 patients with terminal cancer (questionnaire not described in detail). Pain, dysphagia, confusion, and fever were significantly reduced over the first week. Radbruch et al.29 used staff rating to assess 1304 patients referred for palliative care. Self-rating at 2 time points was available from 243 patients (18%), indicating a reduction in the prevalence of patients with moderate or severe levels of many symptoms/problems (e.g., pain, dyspnea, nausea/vomiting and constipation, anxiety, and sadness). Using the McGill Quality of Life Questionnaire, Cohen et al.27 found significant improvement in physical, psychological, and existential problems. Mystakidou et al.6 found a favorable effect over 15 days on all function and symptom scales in the EORTC QLQ-C30 in 120 selected patients with advanced-stage cancer with a KPS value ≥ 50.
The current study is an unrandomized, uncontrolled, longitudinal study, which makes it difficult to make firm statements about effect. Nonetheless, it is relevant to investigate the course of symptomatology: What goes well in daily practice? Which improvements were achieved? Weaknesses aside, several strengths can be found in the current study. First, a relatively large cohort of consecutive patients admitted to the department was examined. Second, the study is longitudinal and reports quantitative, patient-generated data using well validated instruments.
Overall QOL/well-being improved after admission. Pain, nausea/vomiting, insomnia, appetite reduction, and constipation were also reduced. As for fatigue, there was no significant improvement according to the EORTC QLQ-C30. However, improvement was seen after 2 and 3 weeks for 2 subscales of the MFI-20..
HADS depression was unchanged in the study period. This is an example of the problems encountered when evaluating palliative care. The effect of an antidepressive treatment usually begins after 3–4 weeks, and the median survival in our patients was 35 days. The median survival of depressed patients (as measured by the HADS) was shorter than for nondepressed patients (24 vs. 57 days) (data not shown). The resulting attrition may very well contribute to the explanation of the lack of change.
Performing a slope analysis would have had the definite advantage of a single-outcome parameter instead of three as used in the current article. However, it is dubious if one can assume a linear course of symptomatology in the current patient cohort. Therefore, the course of symptomatology was described as differences from levels at T0.
Internal consistency for scales as measured by the Cronbach's α coefficient was high for most scales of the EORTC QLQ-C30, which is comparable to the findings of another study.30
To evaluate the patient-reported outcome of specialized palliative care, one must consider the influence of patient attrition, i.e., that only living patients who had strength and motivation to continue in the study can report change in symptomatology. After 1 week in our study, 5% of participants had died and 85% of the surviving patients had participated. After 2 weeks, 19% of participants had died and 84% of the surviving patients had participated. After 3 weeks, 33% of the participants had died and 81% of the surviving patients had participated. Any change in the symptomatology of nonparticipating patients remains undisclosed. Previously reported results show that patients who did not participate at T0 were older and had lower KPS values than participants.8 In a recent article on newly diagnosed patients with rectal carcinoma, nonparticipation or poor compliance in a QOL study was indicative of a poorer prognosis.31 Therefore, the current results could be biased, which could be a problem in all longitudinal studies of patients with advanced-stage cancer. For example, Cohen et al.27 described favorable outcome of treatment over 7–8 days for 88 patients, acknowledging that these represented only 8% of patients admitted in the study period. The corresponding compliance rates in the current study are 53%, 45%, and 36% for Δ1, Δ2, and Δ3, respectively.
In the palliative setting, poor study compliance is often a sign of poor performance status and/or of pronounced symptomatology. Patients with advanced-stage disease and short survival would tend to comply poorly or not at all. However, the finding that patients who completed fewer questionnaires had significantly higher initial scores for many symptoms does not automatically suggest that the change in symptomatology over time, if measured, would be different from the change in scoring in fully compliant patients. The initial scoring for EORTC QLQ-C30 physical and cognitive function, as well as fatigue and dyspnea (EORTC QLQ-C30 and ESAS), was significantly different between completion levels, indicating that poor compliers had systematically more pronounced initial symptomatology (Table 5). However, testing Δ1 against completion level, no significant differences were found for the change in the self-assessed parameters. Therefore, the evidence does not support an assumption that the effect of palliative care is inferior in patients dropping out, nor that the results of the current study (and similar longitudinal studies) should be biased towards overoptimistic estimates of treatment outcome.
In conclusion, improvement was found mainly for well defined physical symptoms (pain, nausea, lack of appetite, constipation, sleeplessness), as well as for global health status (QOL) and well-being. One would expect the patients' QOL to deteriorate over time without palliative intervention. The findings that longitudinal data exist for the majority of the patients, that significant deterioration was not found for any parameter, and that significant improvement could be detected for several parameters, strongly suggest a broad and positive effect of palliative intervention. The hypothesis that patients with poor compliance with the study had poorer treatment response compared with patients who managed to comply optimally, and thus that incomplete data are biased towards overestimation of positive change, could not be verified for any of the self-assessment scales and items, not even when inspecting for tendencies in the data. Thus, clinicians and researchers should not be discouraged from performing longitudinal symptom assessment in palliative care.
When the study was initiated, there was no consensus on which of several validated questionnaires were the optimal for symptom evaluation in palliative care. Therefore, a number of validated instruments were applied. Currently, consensus has still not been reached, and many different instruments are used to measure QOL. Our study, which uses more than one instrument, helps in the comparison to other studies. There is no doubt that the current assessment approach is quite time consuming for the patients as well as the staff, and it might not be feasible for use in the long run. A reduction in the number of questionnaires and items would probably increase the completeness of data. After the current study, we limited the instruments used to the EORTC QLQ-C30 and HAD scales and a few additional items.
If evaluations using similar design and methodology were used in other palliative care institutions, comparison of treatment outcomes would be very informative. Furthermore, the current study may form the basis of continuous quality improvement by identifying areas where treatment might be improved and selecting topics for intervention studies.
- 1Assessing the outcome of palliative care. Eur J Palliat Care. 1999; 6: 203–205..
- 10Pain characteristics and treatment outcome for advanced cancer patients during the first week of specialized palliative care. J Pain Symptom Manage. 2003; 27: 105–114., , , et al.
- 19EORTC QLQ-C30 scoring manual. 3rd edition. Brussels: EORTC Quality of Life Study Group, 2001., , , , , .
- 20Cognitive impairment. In: RobinsLN, RegierDA, editors. Psychiatric disorders in America, the epidemiologic catchment area study. New York: Macmillan, The Free Press, 1991: 291–327., , , .
- 23Psychometric theory. 2nd edition. New York: McGraw-Hill, 1978..
- 30The European Organization for Research and Treatment of Cancer (EORTC) modular approach to quality of life assessment in oncology: an update. In: SpilkerB, editor. Quality of life and pharmacoeconomics in clinical trials. 2nd edition. Philadelphia: Lippincott-Raven, 1996: 179–189., , , .