Die urpsrüngliche wissenschaftliche Publikation zum LQ-Recorder
Es folgt die Version des Autors, welche sich geringfügig von der publizierten Fassung unterscheiden kann. Tabellen und Grafiken sind hier nicht enthalten.
Practical aspects of quality of life measurement: Designand feasibility study of the Quality-of-Life-Recorder and the standardizedmeasurement of quality of life in an out-patient clinic
Jörg Sigle, Franz Porzsolt Cancer Centre of the University of Ulm,Oberer Eselsberg, 89081 Ulm, Germany
As quality of life (QL) data are of increasing importance, obtainingand processing of these data should be economic as well as quick, easyand standardized. For this purpose, we developed a Quality-of-Life-Recorderwhich is handled by the patients like a conventional paper version of aquestionnaire except that an electronic pen is used. The system computesthe results stores them ready for further processing and can generate aprintout. Questionnaires in any language are supported. A first acceptancetest with a small number of patients led to minor improvements only. Theintegration into a larger study focussing on patients' and staff'sacceptance as well as additional personnel and management requirementswas investigated. Nurses reported that patients had even instructed theirsuccessors how to use the QL-Recorder. Even elderly or slightly handicappedpatients had no difficulties. No additional personnel was required forsystem handling or patients' instruction. This study is not representative,as flaws in patient recruitment resulted in a registration rate of only30% of the 1100 out-patients who had visited the clinic during the studyperiod, although the system users' age distribution resembles thenational demographic curve. In the third phase of the project, we triedto collect representative data from out-patients over a defined period oftime using the EORTC QLQ-C30 (+3). During 19 days, 1315 patients visitedthe out-patient clinic and we managed to track 1267 (worst case, eq. 96,3%).Of the 1157 eligible electronic questionnaires, 1120 had matching entriesin the clinic's diary, representing 94,8% of the desired patientgroup. Their age distribution resembles the national demographic curve.Only 1,1% of the patients declined to participate in the study. The completenessof the questionnaire data exceeded 99,96% (No question was forgotten, butin very rare cases patients said they could not answer one). Descriptivestatistics, reliability and inter-scale-correlation yielded results similarto those from the literature on the validation of the QLQ-C30. We foundno evidence for substantial changes in the test's psychometric parametersinduced by the use with the QL-Recorder. A cost analysis showed that usingthe QL-Recorder, one QL-measurement with guaranteed data quality in onepatient can be done at the total expense of about DM 2.50 (U.S. $ 1,80).For several sub-groups, we compared significant differences in the QL resultsto expectations based on clinical experience. We also performed a clusteranalysis. Our investigations suggested some relationship between differentareas of internal medicine and characteristic changes in QL-measurement'sresults.
Conclusions: Using the QL-Recorder, quality of life can be includedas a routine measure in everyday clinical practice. There is no evidencefor an influence on the psychometric parameters of a questionnaire. Usingthe QL-Recorder, we were able to obtain representative QL data from out-patients.With its open design, simple handling, possibiliterat/lities for integration andits cost efficiency, the QL-Recorder is suitable for ubiquitous use.
Although "Quality of life" (QL) has gained importance duringthe last years, the review of only a few articles on that subject showsthat there is no broad consensus neither about the interpretation of theterm nor about the methods applied for quality of life measurement. Inthe present work we would like to report practical experiences made withthe Quality-of-Life-Recorder we developed. This instrument makes validatedquality of life tests available for use as an everyday standardized routinemeasure. By this work, we would like to support quality of life measurementas a common element in research and practice for investigation and documentationof the success of our measures.
The development of the Quality-of-Life-Recorder included four phases:Construction and acceptance test of the QL-Recorder, investigation of acceptanceand additional personnel and management requirements, collection of representativedata from a large patient sample in a standardized procedure and statisticalanalysis of the data obtained.
As a report on the whole project can be found elsewhere (6), this articlewill focus on phases III and IV of the project. Theoretical aspects ofQL-measurement are discussed elsewhere (1,2,3,7,8).
Collection of representative QL-data
Special requirements in the out-patient clinic
In a university's out-patient clinic, a large number of patientsare distributed to a certain number of physicians and examination faciliterat/lities.Reliably including all patients into a study not only requires enough hardwareequipment but also depends on adequate organisation of patient recruitment.The new, temporary examination has to be integrated smoothly into establishedstructures and processes.
Hard- and software
In a dedicated room located next to the reception desk of the out-patientclinic, we installed six Quality-of-Life-Recorders, using standard PCsin a network under Novell Netware 3.11. Each QL-Recorder ran MS-DOS, GraTaSimV. 2.91 and QLQC-33 V. 1.2. (4, 5, 6)
In phase II, we had found that the completeness of data in a QL-assessmentdepended primarily on the management of the assessment and on the involvedpersonnel's cooperation. It cannot be guaranteed that each patientvisits the desk before each consultation. To cope with this situation,we supplied posters in the whole area of the out-patient clinic asking patientsto visit the QL-measurement prior to each consultation. Nurses and physicianswere asked to send patients who had no printout from the QL-measurement,to the QL-measurement room. There and in the waiting areas we provideda stock of papers for patient information. In the QL-measurement room weprovided an additional stock of 50 copies of the paper version of the QLQ-C30(+3) for patients who would not agree with an electronic assessment butwould fill in a paper questionnaire.
During the whole sampling period, two persons from 8:00 to 10:00 a.m.and a single person during the rest of the day asked incoming patientsto fill in the questionnaire, gave short instructions and stayed nearby,to help if any difficulties should occur. Any problems were documentedas well as patients' reasons for disagreeing in participation. Thelast attendant of a day also performed the measures for quality assurancedescribed below.
The patient's course
Inclusion- and exclusion criteria, patient selection
QL-measurement was defined to be an "examination essential foreach patient who did not meet the exclusion criteria".
Exclusion criteria were: emergency, patient was bedridden or could notvisit the QL-measurement for other reasons, patient did not speak German,visual problems, patient was already hospitalized, patient visited theAutomatic-Implantable-Cardioverter/Defibrillator-clinic only or patientwas a child.
Inpatients would visit the out-patient-area for several special examinations,for example for ultrasonic examination of the thyroid gland. All patientsvisiting the AICD-check were excluded a priori because the responsiblepersonnel stated that the additional QL-measurement would disturb the courseof the sheduled examinations and could not be imposed to the patients.Patients who did not agree to participate in the study were asked to presentthemselves in the QL-measurement room for registration of their initialsand date of birth. As the QLQ-C30 (+3) defines a time-frame of one week,we only scheduled a single measurement for each patient during the shortassessment period of the study.
Chronological integration of the quality-of-life measurement
Most of the patients have to wait for a period of time after they havepresented themselves at the desk. We considered this period of time asoptimal for the QL-measurement as it would not mean additional length ofstay for the patient. When patients entered the out-patient clinic, postersasked them to visit the QL-measurement before their consultation. If theyhad no QL printout, they were sent there from the reception desk, fromthe blood taking room, from several special examinations, or, finally,by the physician after the consultation.
Every patient was asked to complete the QLQ-C30 for study purposes.Participating patients were instructed how to use the QL-Recorder and completedthe electronic version of the EORTC QLQ-C30 (+3). Immediately afterwardsthey received a printout showing patient ID, time and duration of the assessmentand the results of the QLQ-C30 (+3), numerically as well as in a simplegraph. If a patient disagreed to take part in the study, his ID and reasonfor disagreement were recorded. The patient received a short note confirmingthe registration in the QL programme.
The first level of quality assurance depended on all involved personnel:all coworkers was asked to send patients without a QL-printout to the QL-measurement.For the second level, at the end of each day registered patients were comparedto those who had been expected, according to the clinic's diary.If missing patients were reported, the responsible specialty was identified.Together with the involved personnel we checked whether a patient had cancelledhis appointment, or whether he had not been sent to the QL-measurementfacility.
In the third level of quality-assurance all patient identificationsrecorded in the QLQ-files with redundant entries removed were comparedto all entries of the clinic's diary after the end of the study.When this comparison was finished, a several QLQ-files had no matchingentry in the clinic's diaries. Therefore, after several weeks wechecked in the clinic's diary whether matching appointments had beensupplemented in the meantime and repeated the comparison.
Finally, we received information about the number of patients in theout-patient clinic from the clinic's data center. For each department,the number and proportion reported by the data center was compared to thenumber and proportion of registered patients. Moreover, the data centercomputed the total number of patients who visited the out-patient clinicduring the assessment period. We compared this number with the number ofmatched QLQ-files.
A study involving electronic data processing is required to meet substantiallyhigher requirements in data protection than everyday clinical tools. QL-measurementwas located in a separate room which was locked in off-service-hours, orcontinuosly supervised by at least one staff member. Our fileserver waspassword protected and did not have a monitor nor a keyboard for securityreasons. Patients' printouts were treated like other examinationreports. All information was entered directly into the computer. Access-rightswere set to write-only except for the required programmes. As no connectionto other networks existed, information was sealed in our network as soonas the patient ticked the "Finished"-field on the questionnaire.The enhanced access-rights for the daily quality-assurance were passwordprotected and only available in a small time-window from one hour beforeto one hour after the end of service-hours.
A poll concerning the study about 4 weeks after its completion assessedstaff's opinion about additional work load for personnel and patients.Physicians were asked to rate the importance of quality of life. The questionnaireincluded 6 visual-analogue-scales. Results were plotted and examined forcorrelation using Spearman's Rho.
Organizational experiences and cost calculation
We examined the patient flow using automatically recorded data. Resultsmight be important for the design of new QL-faciliterat/lities. We also calculatedthe costs that might be expected for a routine QL-measurement in a patient-populationcomparable to the reported one.
We used only those QLQ-Files for exploratory data analysis which couldbe matched to an entry in the clinic's diary. During the assessmentperiod, a number of students had a medical examination because of theirentrance into their final year internship ("Praktisches Jahr",PJ). They were treated like a control group.
Besides commercially available software (MS Excel and SPSS), softwarewas programmed according to the requirements for the matching of patients,separation of patients outside of SPSS, and for all other special tasks.
We computed some values which have already been used in the validationprocess of the paper version of the QLQ-C30 and compared the results withthose found in the literature to search for obvious influence of the electronicQL-Recorder on the questionnaire results.
Analysis of distances and Cluster-analysis
As a part of the exploratory data analysis, we applied different methodsof hierarchical and k-means-cluster-analysis to check whether or not andhow well different sub-populations could be distinguished.
Cluster analysis is a method which tries to generate clusters of casesof a sample. Cases within a cluster shall be as similar as possible whilethe clusters shall be as different as possible. The first step of hierarchicalcluster analysis is to compute all distances between all patients'results. First generation clusters are made from data-sets located mostclosely together. Consecutively, clusters are being grouped together untilall cases are agglomerated into one cluster. A report of the whole processof clustering can be plotted as a tree diagram which enables the user toidentify especially distinct clusters. A simplified method of cluster analysisis k-means-cluster analysis. Here, the number of clusters to be formedis predetermined. Second, preliminary centers of the final clusters haveto be specified. They are modified during the process of the analysis.
With a sample of over a thousand patients, a hierarchical cluster analysisrequires a great deal of computation time and excessive storage capacity,while the resulting plot is probably difficult to be interpreted. For thisreason, we did not carry out the clustering process completely. Instead,we computed the distances between all multi-dimensional QL-vectors of allpatients. They were examined in a distance histogram. This method was appliedusing vectors of all 18 dimensions or using vectors of a few selected dimensionswhich were expected to distinguish the specialties.
Asking, whether results from patients of the same specialty would besimilar, we finally sorted all distances between all possible pairs ofpatients' QL-vectors (about 624 000) in ascending order. For then closest and the n most distant pairs of QL-vectors, we checked whetherboth patients of the pair belonged to the same specialty.
Characteristics of different specialiterat/lities
To identify differences in QL among out-patients, we compared the resultsof each specialty with a control group. This control group consisted ofall included patients except for the PJ-students and the patients of thisparticular specialty. Observed characteristics of the subgroups were comparedwith expected clinical pictures. For this analysis, we compared the arithmeticmeans of the sub-populations and computed the two-sided significance levelfor the difference of means using the non-parametric Mann-Whitney-U-Test.
Assessment period, number of patients, assessment rates
Patients were included into the Ambu2-study during 19 days. During thisperiod of time, 1315 patients were recorded in the clinic's diary,including appointments that had been entered up to four weeks after theend of the study. Out of these, 101 patients had to be excluded and 33were inpatients. The resulting number of patients in our target group was1181.
During the same period of time, we obtained 1156 electronically completednon-redundant QLQs and one QLQ completed on paper. Out of these, 1120 hada matching entry in the out-patient-clinic's diary. 13 patients hadnot agreed to take part in the study.
In the worst case, completely ignoring the QLQ-files that could notbe reliably matched with entries in the clinic's diary, the resultingregistered proportion of the target group was 1133/1181=95,9% and the proportionof the target group represented by completed questionnaires was 1120/1181=94,8%.
The data of the 1120 patients with matching QLQ-file and entry in theclinic's diary served as data-base for further analysis.
The age distribution of the patients in the study resembles the Germandemographic curve except for very young children (Fig. 1). Children areadmitted from the childrens' hospital to our hospital only for specialdiagnostic procedures. Besides, the QLQ-C30 was not designed for youngchildren.
Compliance and completeness of the questionnaires
Out of 1133 registered patients, only 13 disagreed to take part in thestudy while 1120 agreed. The resulting compliance was 98,9%. Patients whodid not agree to participate, usually not even visited the room of QL-measurement.The most frequent reason was being afraid of delaying one's appointment.An elderly patient was afraid of disadvantages resulting from giving "wrong"answers. One patient who did not want to participate because of a "generalsuspicion against questionnaires", changed his mind and completedall questions very carefully. Half an hour later he came back and wantedto correct some of his answers after he had studied the printout.
The completeness of questionnaires was greater than 99,96%. Missingdata did not result from overlooked questions, but some patients said,they could not answer single questions.
Distrubution of excluded patients
The frequencies of patients meeting the different exclusion criteriaare shown in Table 1. Almoust 70% of the exclusions occured due to languageproblems or because the infrastructure of the AICD-clinic, which they visited,was not compatible with the integration of QL-measurement. As a single patientmay meet more than one criterion, the total number is 114.
Evaluation of the staff poll
The staff poll showed that most staff members thought, they were wellinformed. They thought that QL-assessment was important, that the measurementwas no major burden for the patient and that their own additional workload was rather small. We found that people who thought they were wellinformed usually talked to patients more (r=0,63, p=0,003) and thoughtthat QL-measurement was more important (r=0,44, p=0,045) than those whothought they were not well informed. Interestingly, we found that the physicians'answer "I payed attention to the QL-results in the patient file"correlated with "I think QL-measurement is important" (r=0,75,p=0,048) and correlated negatively with "I got additional work loadfrom the study" (r=-0,72, p=0,055).
Integration of the study in the out-patient clinic's environment
Patient flow varies over the day
The frequency of patient visits to the QL-measurement room versus timeof the day is shown in Figure 2. It shows that there are peak values of25 patients per hour in the morning and a decrease to only 1/5 to 1/10of this flow in the afternoon. This phenomenon influenced the QL-Recorders'system load: while we needed 4 to 6 systems around 8:00, 9:00 and 10:00,a single system was usually sufficient in the afternoon. With an averagecompletion time of 5.5 minutes, only very small delays occured.
Cost and personnel requirements
A cost calculation for 5 years of quality of life measurement is shownin Table 2. Considering 90 000 QL-measurements during a period of fiveyears, we calculated the total costs for a single assessment to be belowUS $2,00.
This calculation does not include costs for planning a study or forevaluation of the results.
Descriptive statistics, reliability and inter-scale-correlations
All EORTC-C30 (+3) dimensions yield results ranging from 0 to 100, butwith different resolutions.
The results of descriptive statistics and reliability for the completesample are shown in Table 3. The results use the complete possiblerange in all dimensions. The quartiles show that dimensions assessing functionare left-skewed while symptoms are right-skewed. Cronbach's a greaterthan 0.7 in most scales and still around 0.6 in KF (0.66), PF (0.62), NV(0.57). Only the RF-scale has an a of 0.36.
For a comparison with literature (1) we computed inter-scale-correlationsaccording to Pearson and found (all p-values<0.005): Correlations betweenfunction scales ranging from 0.27 to 0.7. The strongest correlations werefound between RF and NRF as well as SF and NRF. Correlations between globalmeasures ranged from 0.29 (with symptoms) to 0.65 (with functions) withthe strongest correlations between QL and NQL. Correlations between symptomscales ranged from 0.04 to 0.52, preferring the area around 0.3. The scaleFA showed the strongest correlations with symptoms (around 0.3) as wellas with functions (around 0.6). Obstipation, diarrhea and financial impactonly showed correlations in the range of 0.1 to 0.2.
Analysis of distances and Cluster-analysis
The distances between PJ-students were short even when all 18 dimensionswere included, and in a distance histogram they appeared as a sharp peak.Excluding the PJ-students, other subgroups of patients produced broaderpeaks only when vectors constituent of a small number of dimensions wereused. If additional dimensions were added, these peaks melted into a normaldistribution. For example, a peak resulting mainly from cardiac and pulmonologicpatients could be distinguished well using vectors of age, physical functionand dyspnea, while it disappeared in a distance histogram of vectors ofall 18 dimensions.
These results indicated that the postulated clusters would be distributedover all the space available with large overlapping. This was expectedas the resolution of the dimensions of the QLQ-C30 (+3) is quite low. Becauseof this, in 1120 cases there is a high probability that each sub-population'sresults cover the whole available range in each dimension.
Using several values for n in the range of 10 to 1000, we observed thatin every case there were more pairs of patients from the same specialtyamong the n closest pairs of vectors than among the n most distant pairsof vectors.
As a hierarchical cluster-analysis requires a great deal of computationtime and excessive storage-capacity, and as the results of the groupingprocess in such dimensions were not expected to be interpretable, we didnot run a complete automatic grouping, but performed a k-means-cluster-analysisinstead. This shortened method - depending on the initial values - couldidentify clusters, but it could not generate any clusters exactly resemblinggroups of patients from different specialties.
Search for characteristics of groups sorted by specialties
For some selected specialties, the following diagrams show how theirresults differ from all other patients.
M. Crohn / Coliterat/litis Ulcerosa
This sub population (Fig. 3) differs from other patients by lower averageage, more fatigue, more pain, more appetite loss, more diarrhea. Thesecharacteristics fit the clinical picture of both diseases, i.e. an agepeak between 20 and 40 years, inflammatory disease of the intestine withabdominal pain, diarrhea and fatigue.
This sub population (Fig. 4) differs from other patients by higher averageage around 60 years, worse physical funcion, lower global quality of life,more dyspnea. These findings reproduce the clinical picture of cardiologicpatients.
This sub-population (Fig. 5) differs from all other patients by: moderatelyhigher average age, reduced emotional function, reduced social function,fatigue, appetite loss, obstipation, differences in physical function,dyspnea and pain are not significant, but a larger sample might lead toa better characterization. The reduction of emotional function could beexplained by anxiety caused by episodes of asthma. Appetite loss is a commonsymptom in pulmonal diseases. Obstipation is one of the most obvious differencesbetween pulmonologic and cardiologic patients. It might be explained asa side-effect of the medication for gastric ulcer prophylaxis during corticoid-therapy.
This sub-population (Fig. 6) differs from all other patients by loweraverage age, around 37 years, reduced role function better emotional function,cognitive function and global quality of life, more financial difficulties.The lower average age would be explained sufficiently by the lower transplantationrisk of younger patients which enables to receive this treatment more frequently.Differences in the dimensions EF, CF and QL might be caused by the selectionprocess associated with the procedure of BMT and/or by special emotionalsupport received by these patients.
Experiences made in the course of the present study
Advantages and disadvantages of the used system
Even during the first observations of patients' interaction withthe QL-Recorder, it appeared to be such a simple Man-Machine-Interfacesthat virtually no patient has any problems in its handling. While a patientis filling in a questionnaire, the PC remains completely in the background.The patient focusses on the questionnaire which he is familiar with. Onlyafter the patient has finished the questionnaire, the computer turns upagain: either indicating missing answers, or confirming the completionof the test.
The QL-Recorder has a number of disadvantages as well as advantages.Disadvantages are, for example, the size of the tablet which makes transportationdifficult. It could be substituted by a laptop computer with a digitizerintegrated in it's screen. While multiple page questionnaires requirea turning of pages on the tablett, a screen's contents could changeconsiderably faster. On the other hand, the large size of the tablet guaranteesthat the questionnaire is clearly readable even by patients with visualproblems. The QL-recorder saves the patient from having to learn how tohandle a new data entry tool (like a mouse) and from problems in adaptingto continuosly changing information on a screen. The most obvious advantageis that patients do not feel they are using a computer but that they arecompleting a "paper-like" questionnaire.
Compared to more expensive systems, for example pen-based laptop-computersor personal-digital-assistants with their required hard- and software-infrastructure,the low costs, the use of fully-developed techniques and standard-hardwareand the ability to integrate in current electronic-data-processing-environmentsare other advantages.
Experiences gained from the project
The goal of the project was to integrate QL-measurement in the dailywork of the whole out-patient-clinic with its different subdivisions. Wewanted to assess all patients and we were using a new method. Consequently,cooperation with nursing personnel and physicians was absolutely indispensable.The widely varying patient-flow could be handled very well, and the timerequired for the completion of a questionnaire was about half as long asreported in literature (1) for the paper version. During four weeks, adata set was generated which is representative for our out-patients. Patientcompliance was excellent. The age-distribution of the registered patientsshows that even aged patients could use the QL-Recorder without difficulties.
Results of the staff poll suggest that interest and motivation of theinvolved personnel influence their perception of additional work load,their estimation of stress induced to patients and their estimation ofthe importance of QL-measurement. After all, answers supporting QL-measurementwere predominant, indicating its acceptance by the involved personnel.
During the whole study, no severe problems with the electronic questionnaireitself occured.
These findings demonstrate that the QL-recorder can be integrated intoclinical practice.
Results from further data analysis
The results of descriptive statistics, of reliability analysis and ofinter-scale-correlation resemble the data from the validation process ofthe QLQ-C30 as reported in literature, without beeing completely identical(1). As the observed differences were non-systematic, there was no evidencefor a change in the questionnaires' characteristics induced by theelectronic instrument.
The low Cronbach's a for the RF-scale reproduces the findingsreported in literature. The low value of NV might be influenced by thefact that this symptom was quite rare in our population (average=9,8; 75%reported values below 17). PF and KF approach 0.7 while all other scalesexceed it. and The very high values of 0.89 and 0.91 for the QL and NQLscales might indicate that the people in our population identify qualityof life with their physical condition or general health constitution toa high degree.
The experiences with the new scales NRF and NQL reported on the November1994 meeting of the EORTC Study Group on Quality of Life in Trondheim couldbe reproduced: both scales show a better reliability than their predecessors.
In the comparison between characteristic results of sub-populationsand their clinical pictures, every characteristic result corresponded withthe clinical picture of its sub-population. This finding supports the reliabilityof the questionnaire results. In cluster analysis, the results of pairanalysis of distance vectors support the hypothesis that patients fromthe same specialty have similar results with a higher probability thanpatients from different specialties. Nevertheless, the questionnaire cannotseparate the specialties distinctly throughout all dimensions, as the k-means-clusteranalysis showed. These observations suggest that similar clinical picturesmay cause similar symptoms or functional impairment which results in bettercorrelation among results from patients of the same specialty versus resultsfrom patients of different specialties.
We demonstrated that the Quality-of-Life-Recorder can be used to collectrepresentative QL-data from patients in an out-patient clinic. No evidencewas found for a significant influence of the QL-Recorder on the psychometricparameters of the EORTC QLQ-C30 (+3). QL-measurement was well acceptedby staff and patients; patient compliance was high (98,9%). The QL-Recordercan help to save time, to save money and improves data quality. It provedto be a useful tool for scientific use as well as for daily clinical practice.
Some distinctive differences between QL-results of several sub-populationsof our out-patients were found. The characteristics of each sub-populationcorresponded with its predominating clinical pictures. This finding andresults of cluster analysis supported the hypothesis that although it maybe impossible to identify reliably the underlying disease from a singleQL-measurement, its results have clinical validity and value.
We'd like to thank the following persons and institutions:
M. Fink, B. Mihanovic, G. Lutz, B. Schäffler for taking good careof the patients, Mr. Rettenberger and the staff of the out-patient clinicof the Medizinische Klinik der Universität Ulm for their extremelyfine cooperation, W. Voigt and Mr. Ziesel for additional technical equipment,Dr. Kuhn who provided data from the clinical data center, I. Brand andS. Goldmann for their work as secretaries of the Cancer Centre of the Universityof Ulm, Fa. SPSS GmbH Munich for providing the SPSS-Software in time, A.Coates, W. Gaus, H. Heimpel and D. Machin for their advice in statisticalquestions, W. Streibl for early support of my computing ambitions, A. Piffelfor criticism and suggestions and H. Sigle for improving proofreading,
The reported work is part of a doctoral thesis prepared by cand. med.Dr. med. Jörg M. Sigle at the University of Ulm, Germany under the supervisionof F. Porzsolt.
1 Aaronson NK, Ahmedzai S, Bergman B, Bullinger M, Cull A, Duez NJ,Filiberti A, Flechtner H, Fleishman SB, de Haes JCJM, Kaasa S, Klee M,Osoba D, Razavi D, Rofe PB, Schraub S, Sneeuw K, Sullivan M, Takeda F forthe EORTC Study Group on Quality of Life: The European Organization forResearch and Treatment of Cancer QLQ-C30: A Quality-of-Life Instrumentfor Use in International Clinical Trials in Oncology. Journal of the NationalCancer Institute 1993; 85(5): 365-376
2 Dawes, RM. The robust beauty of improper linear models. Am Psychol34:571-582, 1979
3 Osoba D, Aaronson NK, Till JE: A practical guide for selecting quality-of-lifemeasures in clinical trials and practice. In Osoba D (Hrsg): Effect ofcancer on quality of life. CRC Press, Boca Raton, 1991, 89-104
4 Sigle, JM: SC-JMS GraTaGen and SC-JMS GraTaSim software manuals, JörgM. Sigle, Kunstvolle EDV & Elektronik 1990 - 1995
5 Sigle, JM: SC-JMS GraTaSim / Lebensqualität - Ein angenehmesSystem zur computergestützten Erfassung der Lebensqualität, QL-Recordersoftware manual, Dr. med. Jörg M. Sigle, Kunstvolle EDV & Elektronik 1993- 1995
6 Sigle, JM: Praktische Aspekte der Lebensqualitäts-Messung: Diestandardisierte Messung der Lebensqualität bei Ambulanzpatienten miteinem elektronischen Lebensqualitäts-Recorder, Dissertation zur Erlangungdes Doktorgrades der Medizin der Medizinischen Fakultät der UniversitätUlm, 1995
7 Till, JE: Uses (and some possible abuses) of quality-of-life measures.In Osoba D (Hrsg): Effect of cancer on quality of life. CRC Press, BocaRaton, 1991, 137-154
8 Ware JE and Sherbourne CD: The MOS 36-item Short-Form Health Survey(SF-36). I. Conceptual framework and item selection. Med Care 1992; 30:473-83