Abstracts Listing for ISOQOL Annual Meeting 2001
Current as of 8/16/01
ORAL
1/ COMPARING FIVE METHODS OF UTILITY ASSESSMENT ACROSS DIFFERENT DISEASE STATES
Patrick W. Sullivan, Michael B. Nichol, Denise R. Globe, Pharmaceutical Economics and Policy, University of Southern California, Los Angeles, CA
The objective of this study was to (1) compare five methods of utility assessment across different disease states and (2) to analyze the construct validity of each method. The study population included 6,921 members of Southern California Kaiser Permanente. The HUI Mark II, SF-36 and visual analogue scale (VAS) were directly administered. The Nichol/Sengupta/Globe (NSG) algorithm and the Brazier method (both VAS and standard gamble [SG]) were applied to SF-36 scores to calculate summary utility scores. Dummy variables for 23 different comorbidities were created from AHFS codes. OLS was used to regress utility scores on the 23 comorbidities controlling for sociodemographic variables for all five methods. Generic mean utility scores were also compared. The majority of the dummy variables representing different comorbidities were statistically significant (0.05 level) for each of the five methods. The particular comorbidities that were statistically significant varied by method. For example, for the larger diseases (n>300), the Brazier SG method captured the marginal effect of 13 different diseases, but did not capture the remaining 2 (thyroid replacement and epilepsy). The differences in the parameter estimates and the standard errors between the five methods varied depending on the disease. As an example, the parameter estimates (standard errors) of the marginal impact of Depression on utility for all five methods are as follows: HUI2 -0.08 (.007); NSG -0.08 (.005); Brazier SG -0.04 (.003); Brazier VAS -0.09 (.007) and VAS -0.06 (.007). Generic mean utility scores were: HUI2 (.80), NSG (.80), Brazier SG (.90), Brazier VAS (.65), and VAS (.68). All five methods appear to be valid measures of disease-related morbidity for a variety of comorbidities. The five methods differ in their ability to capture morbidity associated with certain diseases. The Brazier SG method and the HUI2 were substantially different in their estimates of generic utility under uncertainty.
2/ COMPARISON OF THREE MEASURES OF UTILITY WITHIN DIVERSE ETHNIC GROUPS
Denise R. Globe, Joanne Wu, Pharmaceutical Economics and Policy, University of Southern California, Los Angeles, CA
The purpose of this study was to compare health utility scores generated from three different methods. The data set consisted of sociodemographic, pharmacy, and survey data collected from 6,924 members of a large managed care organization. Subjects who completed both the SF-36 and the HUI were included in this study. HUI scores were calculated using the standard algorithm. Two utility score estimates were calculated from the SF-36; (1) a previously published Ordinary Least Squares (OLS) algorithm and (2) the SF-6D. Mean subgroup scores by ethnicity were compared using one-way ANOVA to assess the utility scores between race subgroups. The cohort included Caucasian (59%), African-American (21%), Latino (12%), and Asian (6.8%) subgroups. The HUI was not significantly different across ethnic groups (Caucasians 0.79, African-Americans 0.80, Latinos 0.81, Asians 0.82). Asians had significantly higher unadjusted estimated utility scores from both the OLS model (mean score 0.83) and the SF-6D (mean score 0.92). After adjusting for gender, income, smoking, employment status, and the presence of chronic disease, Caucasians rated utility significantly lower compared to the other ethnic subgroups (mean HUI score range 0.79-0.81). In the two estimated utility scores (mean OLS range 0.80-0.82, mean SF-6D range 0.90-0.91) Caucasian scores were lower compared to the other three ethnic subgroups (P=0.0001). HUI and OLS scores were significantly lower (mean difference 0.10) than SF-6D scores for both the total cohort and within ethnic subgroups. The results of this study suggest that among diverse, English speaking ethnic groups, Caucasians rate health utility lower, regardless of the method used to calculate utilities. Furthermore, mean utility scores differ between the three different methods, irrespective of ethnic group. These findings suggest the need to account for differences in utility scores by ethnicity and utility measurement method when incorporating these values for clinical and policy decisions.
3/ CAN UTILITY-WEIGHTED HRQOL ESTIMATES CAPTURE HEALTH EFFECTS OF QI FOR DEPRESSION?
Cathy Sherbourne, Health Program, RAND, Santa Monica, CA
Utility methods that are responsive to changes in desirable outcomes are needed for cost-effectiveness (CE) analyses and to help in decisions about resource allocation. We evaluate the responsiveness of several different methods that assign utility weights to subsets of SF-36 items to improvements in health resulting from two quality improvement (QI) interventions for depression. The data for these analyses are from a group level, randomized, control trial conducted in 46 primary care clinics in 6 managed care organizations. Clinics were randomized to 1 of 2 QI interventions or usual care. The sample includes 1136 patients with current depressive symptoms and either 12-month, lifetime, or no depressive disorder identified through screening 27,332 consecutive patients. We estimated utility values using 7 different methods (developed by ourselves and other investigators) to convert items from the 12- or 36-item Short Form Health Surveys to utility weights. Utilities elicited directly from patients were also evaluated. We conducted intent-to-treat analyses for each utility measure, specifying linear time trends over the 2 years of follow-up. Several utility-weighted measures showed stronger increases in utility values for patients in one of the interventions, relative to usual care, that were similar in pattern to the improved health effects seen for measures of probable depression and emotional well-being. However, gains in QALYs were small. Directly elicited utility values showed a paradoxical result of lower utility during the first year of the study for intervention patients relative to controls. The results raise cautions about the use of utility measures as outcomes in effectiveness studies for depression. Choice of measure may lead to different conclusions about the benefit and CE of treatment. Utility measures which capture the mental health and non-health outcomes associated with treatment for depression are needed.
4/ CONVERGENT VALIDITY OF SF-36 UTILITIES IN PATIENTS WITH INTERMITTENT CLAUDICATION
Johanna L. Bosch, Decision Analysis and Technology Assessment Group, Massachusetts General Hospital, Harvard Medical School, Boston, MA; Molly T. Beinfeld, Elkan F. Halpern, Scott G. Gazelle, DATA Group, MGH, HMS, Boston, MA
To use SF-36 data for economic evaluation in health care, Brazier et al developed an algorithm to assess preference-based utilities. In this study, we assessed the convergent validity of the SF-36 utilities. The SF-36, Health Utilities Index mark 2 (HUI2) and mark 3 (HUI3), EuroQol EQ-5D, standard gamble, and rating scale were assessed in 87 patients with intermittent claudication before percutaneous revascularization. Societal preference-based SF-36 utilities were calculated based on the algorithm of Brazier et al (which they developed using standard-gamble utilities assessed in the general population of the UK). The utilities and scores were analyzed using paired t-tests and uni-variate linear regression analyses. The mean SF-36 and HUI3 utilities were the same (0.66 vs. 0.66, P=0.92). The mean HUI2 utility, standard-gamble utility, and rating-scale score were significantly higher than the mean SF-36 utility (0.70, 0.91, 0.72, P<0.05); the EQ-5D score was significantly lower (0.57, P<0.05). Uni-variate linear regression analyses demonstrated that the best relationship between the sf-36 utilities and outcomes of the other measures was found between the sf-36 and HUI3 (B-coefficient 0.25, Adjusted R2 0.30) and the poorest relationship between the sf-36 and standard gamble (B-coefficient 0.10, Adjusted R2 0.02). The relationship between the sf-36 and HUI2, EQ-5D, and rating scale were moderate to good (B-coefficient 0.25, 0.21, 0.27, Adjusted R2 0.30, 0.33, 0.26, respectively). In conclusion, SF-36 utilities were very similar to the HUI3 utilities in patients with intermittent claudication. Face validity was demonstrated between the SF-36 and the other measures. These results suggest that the SF-36 utilities estimated with the preference-based algorithm of Brazier et al can be used interchangeably with HUI3 utilities in economic evaluation of new interventions in health care.
5/ ESTIMATION AND VALIDATION OF DERIVED UTILITY SCORES FROM THE SF-36 IN LUNG TRANSPLANT PATIENTS
Francis S. Lobo, Cynthia R. Gross, Barbara Mathees, University of Minnesota, Minneapolis, MN
The purpose of this study is to predict and validate utility scores from the SF-36 in a sample of lung transplant patients and to examine the extent to which the resulting scores correspond to self-reported health. Data including the SF-36, Center for Epidemiologic Studies Depression (CES-D) Scale, Symptom Distress Scale (SDS), Illness Intrusion Scale (IIS), Internal Health Locus of Control (IHLOC) and Health Status (VAS) were collected from 99 lung transplant patients via a mail survey. Methods from two studies (Fryback et. al. and Brazier et. al.) were used to estimate utilities. Correlations and group differences methods were used to validate predicted scores. The Fryback scores ranged from 0.43 to 0.83 (Mean=0.643, SD= 0.086);Brazier scores ranged from 0.12 to 0.92 (Mean= 0.596, SD=0.188). High positive correlations were observed for both scores with the SF-36 PCS and VAS that ranged from 0.61 to 0.83.The SF-36 MCS was highly correlated with the Brazier scores (r=0.65) and moderately correlated with Fryback scores (r=0.32). Moderate positive correlations were observed for both scores with the IHLOC (r=0.47 to 0.51). Moderate to high negative correlations were observed for both scores with the SDS and the IIS (-0.36 to -0.79). On both scores significant differences were found among patients grouped by quartiles of CES-D, SDS, IIS, IHLOC, and VAS and at five levels of general health. The CES-D was highly negatively correlated with the Brazier scores (r=-0.71) as opposed to moderate correlations with the Fryback scores (r=-0.41). The results support evidence of validity for these two derived scores as representative measures of Health Related Quality of Life. The Brazier scores appear to capture mental health aspects like depression and illness intrusion to a larger extent. On the other hand the Fryback scores correlate well with physical health measures like the VAS and general health status. However, there is further need to demonstrate the validity of the use of the derived scores in cost-utility analyses.
6/ CANCER PATIENTS’ PREFERENCES FOR COMMUNICATING QUALITY OF LIFE INFORMATION: A QUALITATIVE STUDY
Michael D. Brundage, Radiation Oncology Research Unit, Queen's University, Kingston, ON, Canada; Anne Leis, Psychology, University of Saskatchewan, Saskatoon, Saskatchewan, Canada; Andrea Bezjak, Radiation Oncology, University of Toronto, Toronto, ON, Canada; Deb Feldman-Stewart, Radiation Oncology Research Unit, Queen's University, Kingston, ON, Canada; Lesley Degner, Nursing, University of Manitoba, Winnipeg, Manitoba, Canada; Karima Velji, Radiation Oncology, University of Toronto, Toronto, ON, Canada; Dongsheng Tu, Joseph Pater, NCIC Clinical Trials Group, Queen's University, Kingston, ON, Canada; Paul Ritvo, Family Physicians Research Section, SCRN, University of Toronto, Toronto, ON, Canada
Quality of life (QOL) is a prevalent outcome measure in clinical trials, but it is not known how to best communicate QOL results to new patients with cancer. Our purpose was to explore cancer patients’ attitudes toward ten visual and written formats for communicating QOL information. Using standardised qualitative methods, six focus groups were held (two groups in each of three cancer Centres) with patients who had completed treatment for cancer 6-24 months earlier. Groups were stratified according to patients’ education. To ensure consistency across Centres, group moderators used the same detailed guide, training video, and props. Three investigators independently coded the resulting (audiotaped) transcripts. Identified themes were then compared between coders, and subsequently between educational strata and between Centres. Participants also rated each of the 10 presentation formats (Likert scale) as to their perceived usefulness. Fourteen men and 19 women with a variety of cancer diagnoses participated; 13 (39%) in the three lower-education groups and 20 (61%) in the three higher-education groups. Three preliminary themes were identified: First, simple formats (simple graphs or written text) were generally preferred to more complex graphical information. Second, patients did not wish to receive QOL information out of context or without explanation. Third, individual patients varied substantially as to which visual format they most preferred. The same format, however, was rated favourably by the highest proportion of participants (85%) in both the high- and low-education groups. Additional themes identified by the groups, and variation in findings between groups, will be presented. Cancer patients desire QOL information, but vary in their preferences for its communication. Simple formats are generally preferred. Further quantitative studies of patients’ preferences are planned.
7/ DOCTOR-PATIENT COMMUNICATION ABOUT HRQL ISSUES IN DAILY CLINICAL ONCOLOGY PRACTICE
Symone B. Detmar, Child Health Division, TNO Prevention and Health, Leiden, The Netherlands; Neil K. Aaronson, Psychosocial Research and Epidemiology, Netherlands Cancer Institute, Amsterdam, The Netherlands
Although improving health-related quality of life (HRQL) is an important goal of palliative cancer treatment, very little is known about actual doctor-patient communication regarding HRQL topics during the treatment period. The aim of this study was to investigate the content of communication regarding HRQL issues between oncologists and their patients, and to identify patient-, physician- and visit-specific factors associated significantly with the discussion of such issues. Discussions between 240 patients and 10 physicians were audiotaped and content-analyzed using the Roter Interactional Analysis System and a substantive content checklist. In addition, patients and physicians were asked to complete a number of questionnaires. Physicians devoted approximately two-third of their conversation to medical/technical issues and one-quarter to HRQL issues. Patients' communication behavior was divided more equally between medical/technical issues (41%) and HRQL topics (48%). Of the independent variables investigated, patients' self-reported HRQL was the most powerful predictor of discussing HRQL issues. Nevertheless, in between 20% and 54% of the consultations in which patients were experiencing serious HRQL problems, no time was devoted to the discussion of those problems. In particular, patients' emotional functioning and fatigue were often left unaddressed. The discussion of HRQL issues was not found to be related significantly to patients' preferences for discussing these topics. In addition, the evaluation of tumor progression was not related to the discussion of HRQL topics. The main conclusion to be drawn from this study was that, despite increasing recognition of the importance of maintaining patients’ HRQL as a goal of palliative treatment, the amount of doctor-patient communication devoted to such issues remains limited, and does not appear to contribute, at least in an explicit sense, to the evaluation of treatment efficacy in daily clinical practice.
8/ PHYSICIAN’S COMMUNICATION STYLE, CONSULTATION TIME AND PATIENT’S PARTICIPATION IN ONCOLOGY SETTINGS
Tomoko Takayama, Yoshihiko Yamazaki, Health Sociology, The University of Tokyo, Bunkyo-ku, Tokyo, Japan; Atsushi Fukuuchi, Tsunehiro Nishi, Mitsui Memorial Hospital, Chiyoda-ku, Tokyo, Japan; Masakazu Toi, Katsumasa Kuroi, Kazuo Hayashi, Tokyo Metropolitan Komagome Hospital, Bunkyo-ku, Tokyo, Japan
Patient’s participation has been recognized as important to the patient’s quality of care and quality of life. In this study, we examined the effects of the physician’s communication style and visit-specific factors on the patient’s participation during the consultation. 86 breast cancer consultations were audiotaped and analyzed using the Roter Interaction Analysis System. We employed dependent variables as follows: three observational measures of patient’s communication behavior (patient’s biomedical information-giving, psychosocial information-giving, question-asking) and patient’s perception of the participation. Independent variables: three observational measures of physician’s communication behavior (information-giving, attentive listening, question-asking), physician’s collaborative communication style perceived by patient, and the visit-specific factors (consultation length, patient’s perceived time pressure). The limited duration of the encounter suppressed the patient’s communication behavior, psychosocial information-giving (p<0.05) and especially question-asking (p<0.0001). Patient’s biomedical information-giving and perceived participation increased when the physician’s attentive listening or question asking was accompanied by the patient’s perception of physician’s collaborative style. When the patient felt the consultation was time pressured, on one hand patient’s psychosocial information-giving and question-asking were suppressed, on the contrary, patient’s perceived participation was enhanced. This study suggests that the effects of the physician’s communication style and the consultation length on the particular patient’s communication behavior. Also, the patient’s perceived time pressure not only influences patient’s communication behavior but also distorts the patient’s sense of participation.
9/ MEASURING QUALITY OF LIFE IN CLINICAL PRACTICE IMPROVED PATIENT SATISFACTION
Fabienne Empereur, Epidemiologie Evaluation Cliniques, Chu de Nancy, Nancy, France; Emmanuel Desandes, Recherche Clinique, Centre Alexis Vautrin, Vandoeuvre-les-Nancy, France; Francis Guillemin, Isabelle Leonard, Sylvie Klein, Serge Briancon, Epidemiologie Evaluation Cliniques, Chu de Nancy, Nancy, France
Quality of life measurements use is widespread in the field of clinical and epidemiological research, especially for the assessment of therapy. The value of these questionnaires for individual medical decision-making is yet a matter of debate. The aim was to investigate whether the systematic measurement of QoL at admission, and whether scores being returned or not to the care staff modify patients QoL, care management and satisfaction with care. The CliniQualVie randomized controlled trial was designed as a 2*2 factorial design and a control group : having not QoL measures at admission (control group), having Qol measures at admission and returning results to physicians (Bp group), and returning results to nurses (Bn group), and returning results to both (Bpn group), and no returning results (B0 group) Fifteen days after discharge, all included inpatients received and returned the same type QoL questionnaires and the Patient Judgments of Hospital Quality questionnaire (PJHQ). Two thousand seven hundred and four inpatients have been included since April 1997 to April 2000 at the Nancy University Hospital Center (France) (1/5 per group). B0 group patients have better satisfaction scores (0-100 scale) than control group patients, for several dimensions: "admission" (71.2 Vs 67.2, p<0.01)," hospital environment and ancillary staff" (67.2 Vs 64.9, p<0.05), "recommendations/intentions" (82.6 Vs 79.4 ; p<0.05). In all inpatients who completed QoL questionnaires at admission, there was no significant differences in all satisfaction dimensions between B0, Bp, Bn and Bpn groups. Complete a QoL questionnaire seems to improve satisfaction level in few dimensions, but the provision of feedback on patient's quality of life to health professionals seems to have no effect on satisfaction level. Work supported by a PHRC 1996 grant.
10/ THE IMPACT OF SHARED TREATMENT DECISION-MAKING ON QUALITY OF LIFE IN WOMEN WITH BREAST CANCER
Ann F. Chou, Health Services and Policy Analysis, Joan R. Bloom, School of Public Health, University of California at Berkeley, Berkeley, CA
Biomedical advances have offered women with breast cancer a variety of treatment options that are comparable in health outcomes. Moreover, there is a recent shift toward a view of shared decision-making, which encourages the individual to be an informed patient and the physician to be a respectful provider. This process may be determined by individual beliefs towards participation, the amount of information received and processed, relationship with the physician, the environment in which the decision process occurs, or knowledge of treatment choices. Nevertheless, literature has been limited on this type of decision-making model and its impact. The objective of this study is to understand the impact of the decision-making process on survival and patient’s quality of life. The study sample included 722 women diagnosed with breast cancer between 1994-1997. Descriptive statistics show that women in this cohort were mostly married (80%), White (70%), had some college education (80%), and diagnosed at age 50 or younger. The distribution of disease stage was 14% in-situ cancer, 46% local, 37% regional, and 2% remote. Excluding the in-situ cases, 47% of the women who received surgical treatment had a lumpectomy and 53% had a mastectomy. Fifty-four percent of the women reported having undergone radiation therapy, 66% had chemotherapy, and 29% used Tamoxifen. The Cox proportional hazards model was used to examine the impact of decision-making involvement on survival and the multiple regression analysis is used to determine the impact on quality of life. Overall, 20 % of women left the decision to their physicians, 29% shared the decision-making, 43% made the decision with physician input, and 8 % reported making the decision on their own. The mortality rate for this cohort was 8%. Those who reported having a good understanding of cancer and treatment had significantly better mental heath outcomes. Controlling for demographic factors and disease stage, preliminary results suggest that participation in shared decision-making is correlated with better mental health but poor physical health. On the other hand, decision-making independent of physician input is correlated with poor mental health but better physical health. Overall, different levels of patient participation in decision-making had a varying effect on quality of life. Understanding the effects of participation in medical treatment decision-making is important for designing educational interventions for patients and providers, and updating hospital/clinic policies to include patients in medical treatment decisions.
11/ EVALUATING THE LATENT DISTRIBUTION IN IRT MODELS OF HRQOL DATA
Jakob B. Bjorner, Department of Work Psychology and Sociology, National Institute of Occupational Health, Copenhagen, Denmark
IRT analyses of questionnaire data concerning HRQOL must frequently deal with the problem of skewed data distributions. This may reflect floor and ceiling problems of the items or scales and skewness in the latent distribution. In case of a non-normal distribution of the latent variable, item parameters can still be estimated using conditional maximum likelihood estimation (Rasch type models) or discrete empirical approximation of the latent distribution and marginal maximum likelihood estimation (other IRT models). Floor and ceiling problems in the data can create problems for statistical analyses like regression analyses or analyses of variance even if IRT-based scores (i.e. estimates of the latent variable) are used as the outcome variable. One solution is to use models that combine the IRT model with the regression model or the analysis of variance model. In such a latent regression analysis the dependent variable is the latent variable itself and not an estimate hereof. However, since such techniques assume a conditional normal distribution of the latent variable we still need to evaluate the latent distribution. We show how latent regression analysis can be applied in research on HRQOL and we compare two approaches for examining the latent distribution: (1) examining the empirical (estimated) latent density as implemented in Parscale/Bilog, and (2) comparing the observed and expected distribution on a simple sum scale of the items (building on work by Thissen et al 1995). We use data from HRQOL research and simulated data. We show that in case of floor or ceiling problems, the estimated latent density can be non-normal even when the true latent density is normal. Comparison of the observed and expected sum scale distribution appears promising for testing assumptions about the latent distribution.
12/ CAN THE CAMCOG MEASURE OF COGNITIVE STATUS BE MODELLED USING ITEM RESPONSE THEORY?
Rob J. de Haan, Rebecca Holman, Robert Lindeboom, B. Schmand, Clinical Epidemiology and Biostatistics, Academic Medical Center, Amsterdam, The Netherlands
When considering the quality of life of those suffering from dementia, it is important to examine their cognitive status. This can be measured using one of many multi-item instruments. One of the best known is the CAMCOG. It has been used as a screening instrument and in numerous scientific investigations, having been validated in many populations. In spite of its widespread use, in-depth psychometric consideration of the CAMCOG has not often been carried out. This paper will examine the patterns of responses to the items in the CAMCOG. Firstly, a brief classical psychometric (sum score) analysis will be presented. We will then consider whether more information about the psychometric properties of the CAMCOG can be obtained using some simple graphical techniques. These will lead onto the examination of how the responses to the items can be placed on a single, hierarchical scale with linear measurement properties using the powerful techniques of item response theory. The classical psychometric analysis shows an instrument with a high internal consistency. However, when the response patterns on individual items were compared to the sum scores it became apparent that a large number of the items in the CAMCOG demonstrated an obvious ceiling effect. These items provide very little information on the cognitive ability of patients. The item response theory analysis highlighted a number of items with strange, often random, behaviour in relation to the disease stage of patients. These items can give misleading results when used to measure the cognitive ability of patients. It has been possible to place approximately one half of the 60 items in the CAMCOG on a unidimensional linear scale using the Rasch model. It is hoped that the results in this paper will encourage researchers to explore the use of item response theory in the analysis of data resulting from multi-item instruments.
13/ COMPARISON OF THREE DIFFERENT STATISTICAL ANALYSIS STRATEGIES FOR ANALYZING SEATTLE ANGINA QUESTIONNAIRE QUALITY OF LIFE DATA
Colleen Norris, APPROACH Project, Edmonton, AB, Canada; W. Ghali, University of Calgary, Calgary, AB, Canada; D. Saunders, University of Alberta, Edmonton, AB, Canada; R. Brant, University of Calgary, Calgary, AB, Canada; D. Galbraith, Foothills Hospital, Calgary, AB, Canada; P. Faris, University of Calgary, Calgary, AB, Canada; M. Knudtson, Foothills Hospital, Calgary, AB, Canada
The Seattle Angina Questionnaire (SAQ) is being used with increasing frequency in clinical research to address the quality of life (QOL) outcomes of patients with coronary artery disease (CAD). Results of a review that systematically identified all published studies analyzing SAQ scores, demonstrated that inappropriate analysis methods are commonly used to analyze SAQ scores. The purpose of this paper was to compare three different statistical analysis strategies for analyzing skewed SAQ QOL data. Using the Physical limitation scale of the SAQ, three regression models were constructed. Two logistic models were fitted using binary outcomes and one ordinal regression model was fitted using an ordinal scale as the outcome variable. Comparison of the 3 models demonstrated that the ordinal regression model is more sensitive to the characteristics, specifically the ordinality, of the SAQ data. The ordinal regression model generated estimates of effect that consistently yielded the narrowest 95% confidence intervals. More importantly, the results (summary odds ratios) are generalizable to the entire patient population under study, as the estimated proportional odds ratio is not based on a particular dichotomization of the outcome variable. Although this paper focuses on the SAQ, the issue of the most appropriate statistical analysis for disease specific QOL scales may not be unique to this particular scale. Ordinal scales (particularly common in QOL instruments) even when transformed, risk generating skewed results. The ‘average’ scores for patients with chronic diseases may be concentrated at the top of the scale (ceiling effect) or the bottom of the scale (floor effect) severely limiting the range of scores possible. This makes it difficult using parametric methods to describe differences in QOL post-treatment or changes in QOL over time. The results of this study suggest that analysis of the SAQ should be done using ordinal regression modeling.
14/ THE AMSTERDAM LINEAR DISABILITY SCORE PROJECT (ALDS): DEVELOPMENT AND CALIBRATION RESULTS OF AN ADL-ITEMBANK
Robert Lindeboom, Rebecca Holman, M. Vermeulen, Rob de Haan, Clinical Epidemiology and Biostatistics, Academic Medical Centre, KEB, Amsterdam, The Netherlands
The functional status or QoL of patients is frequently expressed as a total score obtained by summing the (weighted) individual items in a scale. The disadvantage of this approach is that all items have to be presented to patients regardless of their functional level, cross-instrument comparability is not assured and the scores obtained are ordinal rather than linear. Also, because of the correlational methods used in their development, the items in many functional status or QoL scales cluster in the mid-range of the construct of interest. Hence, the extremes of the scales tend to be poorly defined and are obscured by floor and ceiling effects. The goal of the ALDS-project is to calibrate an Activities of Daily Living (ADL) item bank using IRT statistical models that claim to correct for these problems. The item bank consists of 190 items selected from a pool of more than 110 existing ADL scales and covers a large range of the construct. We will discuss the theoretical framework used in the item selection and the specific methods to put all 190 items on a common difficulty scale. The calibration estimates of the more challenging half of the item bank containing the more difficult ADL item tasks will be presented. The estimates were obtained by applying Rasch’s IRT model to the responses of about 1000 outpatients with a range of chronic conditions
15/ THE NOTTINGHAM HEALTH PROFILE IN POPULATION RESEARCH: FLOOR EFFECTS PERSIST IN GROUPS WITH IMPAIRED HEALTH
Sonja Boehmer, Thomas Kohlmann, Institute for Social Medicine, Medical University of Luebeck, Luebeck, Germany
The Nottingham Health Profile (NHP) is known to produce strong floor effects (i.e. percentage of "zero scorers") when used in surveys of the general population. Recently, in a representative German population sample up to 83 percent of the respondents showed zero scale scores on the NHP. The purpose of this study was to investigate if floor effects decrease substantially in population subgroups when only persons with at least moderate health problems are included. Mail survey data from a random sample consisting of N=600 residents of Luebeck, North Germany, aged between 50 and 69 was used for this analysis (51 percent male, response rate 61 percent). Floor effects were calculated for the total sample and for respondents reporting (A) fair or poor general health (26 pct), (B) moderate, severe or very severe pain (33 pct) or (C) depressed mood (CES-D score >23; 5 pct). In the total population sample score distributions of the NHP subscales were highly skewed in the direction of positive well-being with the following floor effects: Energy 72 pct, Pain 59 pct, Emotional Reaction 69 pct, Sleep 51 pct, Social Isolation 86 pct and Physical Mobility 55 pct. In all three subgroups (A-C) these floor effects decreased. In group A and B, smallest floor effects were observed for Physical Mobility and Pain, respectively (A: 14 and 20 pct; B: 19 and 17 pct). In group C, floor effects were markedly reduced in three subscales: Sleep (0 pct), Emotional Reaction (3 pct) and Energy (3 pct). Like in the total population sample Social Isolation produced the highest rate of zero scorers in each of the three subgroups (A: 70 pct; B: 76 pct; C: 25 pct). The study showed that floor effects of the NHP decrease in population subgroups when only persons with at least moderate health problems are included. Yet, this reduction has not been substantial in all scales. According to these results the NHP seems to be suitable as a survey tool only in population subgroups with relevant health problems.
16/ USING ITEM RESPONSE THEORY TO SHORTEN THE LENGTH OF SCALES IN A QUESTIONNAIRE FOR PALLIATIVE CARE. AN EORTC QUALITY OF LIFE GROUP STUDY
Mogens Groenvold, Morten A. Petersen, Department of Palliative Medicine, Bispebjerg Hospital, Copenhagen NV, Denmark; Jakob B. Bjorner, National Institute of Occupational Health, Copenhagen OE, Denmark
When evaluating physical and psychological symptoms experienced by patients in palliative care, the questionnaire should be short but still satisfy standard requirements regarding content validity and psychometric properties. The EORTC QLQ-C30 was judged to satisfy such standard requirements and has the advantage of being widely used. However, ideally, a questionnaire for palliative care should be shorter than 30 items. This project will reduce the QLQ-C30 from 30 to approximately 20 items relevant for patients in palliative care. Results obtained with the shortened version should be directly comparable with scores on the (full-length) QLQ-C30. Two approaches are used: (1) Shortening multi-item scales using Item Response Theory (IRT). (2) Interviews with patients and professionals across Europe to identify items or scales that can be eliminated. We here present results from the first approach. A database containing QLQ-C30 data from 10,815 European cancer patients was created. We developed an IRT model, used it to select items for shortened scales and to predict scores on the original scales from the shortened scales, examined agreement between the shortened scale and the original scale, and compared their measurement abilities. We have analysed the 4-item Emotional Function (EF) scale and the Fatigue scale. The EF scale was shortened from 4 to 2 items. There was very high agreement between EF scores obtained with the original 4-item scale and scores predicted from the shortened 2-item EF scale: the mean deviation was 1.00 on a 100 point scale, the correlation was 0.94. Judged from 'known group comparisons' in 24 data sets, there was virtually no loss of statistical power despite the use of only 2 items. The Fatigue scale was shortened from 3 to 2 items which resulted in a small loss of statistical power for this scale. The surprisingly good preservation of measurement power in shortened scales may be caused by (1) the use of a prediction system (predicting the original scale scores) rather than just estimating a 2-item sum score from the shorter scale, and (2) the omission - following the IRT analyses - of a relatively uninformative item from the EF scale.
17/ A COMPARISON OF ANALYSIS METHODS FOR HEALTH-RELATED QUALITY-OF-LIFE (HRQL) OUTCOMES OF CLINICAL TRIALS
David Osoba, QOL Consulting, West Vancouver, BC, Canada; Grazyna Lieberman, Biostatistcs, Genentech, Inc., South San Francisco, CA
The purpose of this study was to determine differences between three methods of longitudinal analysis of HRQL data: means of scores, means of changes in scores from baseline, and proportions of patients with improvement or deterioration (success/failure rate). These methods were assessed in a randomized trial of trastuzumab (Herceptin) (H) and chemotherapy (C) vs C alone in 469 women with HER2-overexpressing, metastatic breast cancer (Slamon DJ et al. NEJM 2001;344:783). HRQL was assessed in 400 women using the EORTC Quality of Life Questionnaire (QLQ-C30, version 1.0) at baseline, 8, 20, 32, 44 and 56 weeks of treatment. The data for 6 preselected domains (global QL, physical, role, social and emotional functioning, and fatigue) were analyzed. Mean scores (for women still on study compared to the mean baseline score for the entire group) and mean change scores from baseline (the difference at each time point for women remaining on study minus their own baseline scores) were determined by repeated measures ANOVA. Success/failure rates were assessed by calculating the proportions of women whose scores either improved or deteriorated by =/>10 points for at least 12 weeks (Osoba D et al. JCO 1998;16:139). Mean scores, as compared to baseline, improved over time to a similar extent in all 6 domains (e.g., global QL in the H+C group = 66.8 at week 44 vs 59.3 at baseline; in the C group = 62.9 at week 44 vs 58.4 at baseline). However, the means of changes in scores from baseline showed statistically significant improvement (P<0.05) only in the global QL, physical functioning and fatigue domains in the H+C group but not in the C group. The success rate was statistically significantly higher (P<0.05) for global QL and fatigue in the HC group than in the C group. The failure rate was similar for all domains in both treatment groups. In conclusion, calculation of the mean scores over time is biased by a "survivor effect" in patients remaining on study, regardless of the treatment they receive. Mean change scores and success/failure rates present a more accurate picture of HRQL outcomes. (Supported by Genentech, Inc.)
18/ THE GOAL INTERFERENCE SCALE: EVALUATION OF A NEW INDIVIDUALIZED HEALTH OUTCOME MEASURE
Amy Peterman, Ph.D., David Cella, Ph.D., Marianne Brady, Ph.D., Northwestern University, Evanston, IL
Despite the proliferation of health-related quality of life measures, most do not provide information about individually meaningful problems that are amenable to intervention. One reason for this may be that most do not focus upon personally relevant goals which vary across people and which are differentially affected by illness/treatment. Existing literature and patient input were utilized to create the Goal Interference Scale (GIS), a questionnaire that incorporates both idiographic and nomothetic information. Thus, it provides a method for capturing the highly individual impact of illness/treatment while still allowing for comparisons of scores across groups of patients. Thirty-three cancer patients who were currently receiving chemotherapy completed the GIS, the FACT-G (Cella, et al.,1993) and the FACIT-Fatigue Scale (Yellen, et al.,1997)at an initial assessment (T1) and 28 (85%) subjects also completed a 2 month follow-up (T2). The sample was 75% female: median age was 61 years. Descriptive statistics demonstrated a wide range of scores (T1: 8-77; T2: 10.5 – 79), with a possible range of 0 – 100. Subjects experienced a moderate degree of interference at both T1 (mean = 49.8; SD=17.2) and T2 (mean=47.1; SD = 18.3). The scale demonstrated good internal consistency (Cronbach’s alpha = .87 at T1 and .83 at T2). Concurrent validity was demonstrated by significant correlations between the GIS and the total FACT-G, functional well-being, fatigue and global QOL (r’s from -.46 to -.78, p’s<.001). In addition, subjects indicated that the GIS is easy to complete and asks about meaningful and personally relevant issues. Initial testing of the GIS indicates promising psychometric properties and the likelihood that goal interference is a meaningful concept for cancer patients. Further evaluation of the scale’s validity and clinical relevance is ongoing.
19/ COMPARISON OF THE HEALTH UTILITY INDEX MARK III (HUI-III) AND THE SELF-ASSESSED HEALTH STATUS IN THE CANADIAN POPULATION.
Julie Bernier, Jean-Marie Berthelot, Health Analysis and Modeling Group, Statistics Canada, Ottawa, ON, Canada; Ivan Barofsky, The Johns Hopkins University, Baltimore, MD; Pennifer Erickson, Pensylvania State University, State College, PA
The objective of this presentation is to compare a population based multiple domain measure of health, the HUI-III, with the single item self-assessed health status measure (SAHS). The household component (n=17,626 individuals) of the 1994 National Population health Survey (NPHS) was designed to be representative of the Canadian household population. Both the HUI-III and the SAHS have been measured with a variety of socio-demographic (e.g., sex, age, language) and health-related (e.g., diabetes, high blood pressure, restriction of activities) variables. In order to have both the HUI-III and SAHS evaluated on a 0 to 1 scale, the categorical responses to the SAHS question have been transformed into the values provided by Stewart and Ware. Using transformed data, the difference between the HUI-III and SAHS was calculated and studied within socio-demographic groups and for individual having specific health conditions. In order to account for the strong relationship between the socio-demographic and health variables, a multiple regression model was estimated to provide an overall picture of the sources of the observed differences between HUI-III and SAHS. Comparing HUI-III and SAHS in the total sample revealed a mean difference of only 0.10. However, this difference varies greatly with age and level of SAHS. The mean difference can be as large as 0.44 for people aged 25 to 34 who self-report fair health status. In general, the difference is higher for people reporting fair and poor health, aged between 55 and 74 or reporting diabetes, high blood pressure or asthma. The two measures have been shown to be different at the individual level and analysis of these differences reveals information about how individuals respond to different types of items within a questionnaire. In addition, the magnitude of these differences varies as a function of personal characteristics.
20/ HEMOPHILIA: CONSTRUCT VALIDITY OF THE HEALTH UTILITIES INDEX
William J. Furlong, Department of Clinical Epidemiology and Biostatistic, Ronald D. Barr, Department of Pediatrics, McMaster University, Hamilton, ON, Canada; Mahassen Saleh, Department of Pediatrics, Hamilton Health Sciences Corporation, Hamilton, ON, Canada; John Horsman, Department of Clinical Epidemiology and Biostatistic, McMaster University, Hamilton, ON, Canada; Julia Sek, Mohan Pai, Department of Pediatrics, Hamilton Health Sciences Corporation, Hamilton, ON, Canada; Irwin Walker, McMaster University Medical Centre, Hamilton Health Sciences Corporation, Hamilton, ON, Canada
This population-based study assessed the comprehensive health status and health-related quality of life (HRQL) experienced by males age >13 years, with a diagnosis of hemophilia A or B, and active in a geographically-defined health care program. Diagnostic variables included virologic status and disease severity. A mail survey, using a standard Health Utilities Index Mark 2 (HUI2) and Mark 3 (HUI3) self-complete questionnaire, collected information from patients about their health status during the previous 4 weeks. The HUI2 and HUI3 are generic systems for measuring health status and HRQL. A priori hypotheses included reduced health status being associated with increasing severity of disease, hepatitis B and C (B/C) sero-conversion, and human immunodeficiency virus (HIV) infection. The survey response rate was 101 of 115 (88%). Non-response was due to refusal (n=10), terminally ill from AIDS (n=2), inability to communicate in English (n=1), unable to contact (n=1). Patients' HRQL mean HUI2 score was 0.81 (range 0.20 to 1.00) and HUI3 was 0.71 (range -0.23 to 1.00). There were statistically significant (p<0.05) and important size (>0.10 units) differences in mean overall HRQL scores within the following pairs of patient groups: Pair I - mild disease (n=41) versus severe disease & HIV+ (n=23); Pair II - moderate disease (n=17) versus severe disease & HIV+ (n=23); Pair III - severe disease & HIV- (n=20) versus severe disease & HIV+ (n=23); Pair IV - moderate disease & HIV- & B/C- (n=7) versus moderate disease & HIV- & B/C+ (n=8). Within Pairs I and II, there were significant differences in mean single-attribute utility scores for HUI2 mobility, HUI2 pain, HUI3 ambulation and HUI3 pain. Within Pair III there was a significant difference in mean single-attribute scores for HUI2 pain. The significant differences between clinically important groups were in the expected directions and provide evidence of construct validity for HUI.
21/ ONE YEAR OUTCOMES OF HEMODIALYSIS AND PERITONEAL DIALYSIS IN AN INCIDENT COHORT OF PATIENTS WITH END STAGE RENAL DISEASE (ESRD)
Albert W. Wu, Health Policy and Management, Jane Marsh, Nancy E. Fink, Department of Epidemiology, Kraig Kinchen, Neil R. Powe, Department of Medicine, Johns Hopkins University, Baltimore, MD
After nearly three decades of Medicare coverage of dialysis care, the relative outcomes of hemodialysis (HD) and peritoneal dialysis (PD) are not known. A few cross-sectional studies have suggested better health related quality of life (HRQOL) on PD. We examined HRQOL and survival in an incident cohort of patients with ESRD. Patients from over 80 dialysis clinics across the US were enrolled in Choices for Healthy Outcomes in ESRD (CHOICE) between 1995 and 1998. Patients completed the CHOICE Health Evaluation Questionnaire (CHEQ) which includes the SF-36 and 14 ESRD specific domains (Cronbachs alph > .67), at baseline and 12 months later. Dependent variables included HRQOL, survival and renal transplant. Independent variables included age, gender, race, education and comorbidity score. Analyses examined unadjusted and adjusted differences in changes in HRQOL scores, and change (worse, same, better) in overall health status in which death or increase in comorbidity = worse, transplant or decrease in comorbidity =better, and HRQOL was categorized based on +2 SEM. Of 928 patients who completed the baseline CHEQ, 585 also completed the 12 month CHEQ; 101 had died, and 55 had renal transplant. PD patients were slightly younger, more likely to be Caucasian, well-educated, employed and had less comorbidity. Unadjusted baseline scores showed better HRQOL for PD patients (physical function, pain, role emotional, travel, diet, and dialysis access (P<.05)). At 1 year SF-36 scores improved, while ESRD specific domains were mixed. There were relative improvements for HD in SF-36 (physical, general) and mixed results for ESRD domains (PD better for finances, HD better for sleep, unadjusted and adjusted). No differences between HD and PD in overall health status were found. In conclusion, in one of the first longitudinal studies to compare HD and PD patients' HRQOL, we found substantially similar outcomes for the two treatments. General health status improved more for HD patients. However, for ESRD specific domains, there appeared to be trade-offs, with PD better in some, and HD better in other domains. In advising patients about which modality to choose, clinicians should discuss the individual's preferences for specific aspects of HRQOL.
22/ USE OF ERYTHROPOIETIN IN EARLY ANAEMIA IMPROVES HRQOL IN PREDIALYSIS CHRONIC RENAL FAILURE PATIENTS
M. S. Salek, Centre for Socioeconomic Research, Cardiff University, Cardiff, UK; N. Pratheepawanit, Faculty of Pharmaceutical Sciences, Khon Kaen University, Khon Kaen, Thailand
Predialysis patients are usually allowed to become anaemic (Hb greater than or equal 9.0 g/dl) before the commencement of erythropoietin therapy. Several studies have demonstrated the physiological benefits of correction of anaemia in predialysis patients. However its HRQOL benefits have not been established. An open label, prospective, randomised, multicentre study was carried out to assess the impact of maintaining a haemoglobin level of 11.0 +/- 1.0 g/dl on HRQOL in predialysis chronic renal failure patients. Patients who met the inclusion/exclusion criteria were randomised into two groups. In group A, patients commenced erythropoietin therapy at an early stage in the development of their anemia in order to maintain a target hemoglobin level of 11.0 +/- 1.0 g/dl. Hemoglobin levels in group B patients were allowed to fall below 9.0 g/dl prior to receiving erythropoietin therapy. The HRQOL was measured at baseline and a year later using the Renal Quality of Life Profile (RQLP). This is a 34-item questionnaire consisting of 4 domains: eating and drinking; physical activity; leisure activity; and daily life. A total of 65 patients completed the study; 26 (40%) patients were in group A and the rest (n = 39) in group B with similar demographic characteristics. Most patients were male (group A = 81%, group B = 68%) with mean age of 57 (group A = 60, group B = 55), ranging from 28 to 78 years, and the majority stated that, at baseline, they were still working and led a normal life (group A = 73%, group B = 78%). There were no significant differences in baseline RQLP scores between group A and B. Results at one year showed a deterioration of patients’ quality of life in both groups. However, slightly more impaired quality of life was found in group B than group A, except physical activities where group A showed an improvement. In comparison, the difference of quality of life scores in physical activities was statistically significant between both groups (p = 0.03). The findings suggest that maintaining hemoglobin level of 11.0 +/- 1.0 g/dl can improve patients’ physical activities and delay the deterioration of patients’ overall HRQOL.
23/ ORODENTAL HEALTH, ACUTE AND LATE MORBIDITIES, AND QUALITY OF LIFE IN IRRADIATED INDIAN HEAD AND NECK CANCER PATIENTS
Manishi Bansal, Bidhu Mohanti, Naseem Shah, Rama Chaudhry, Sudhir Bahadur, Nootan Shukla Institute Rotary Cancer Hospital, AIIMS, Ansari Nagar New Delhi, India
Approximately 50%-70% of head and neck cancer (H&N Ca) patients receive radiation therapy in India. Assessment of orodental status, radiation-related morbidities and impact of irradiation on quality of life is not consistently carried out and this study has evaluated the above issues in the Indian patients prospectively. Radiation therapy (RT) was delivered to 45 H & N Ca patients (70 Gy, Group A) and 20 patients of cancer other than H&N Ca (46-60Gy, Group B). Orodental health, dental care and oral flora culture, acute and late radiation morbidites (RTOG and LENT/SOMA systems), and quality of life scales (EORTC QLQ-C30) were recorded in both groups. Temporal assessments were carried out in 5 phases-Pre-RT (phase I), during 4th week of RT course (phase II), at end of radiation (phase III), after 1 month of RT (phase IV) and after 4 months of RT (phase V). The caries incidence increased from 0.37% in phase II to 2.8% at phase V(p<0.0001). Periodontal disease recorded in 51% patients before RT increased to 86.6% at end of therapy (p<0.001). Candidiasis was detected in 24(53.3%) H & N Ca patients in phase II. Acute radiation morbidites of head and neck cancer patients were prominent in phase III, affecting skin, mucosa and salivary gland in 15.5%, 66.6% and 97% of patients respectively. Late radiation effects were observed in salivary gland (90.6%) and teeth (58.3%) at 4 months after RT course. QOL assessment showed decline in scores for H & N Ca patients in physical and social function (p<0.0001),and emotional function (p<0.001) scales, for phase I vs IV. The pre-RT global health status was higher in H&N Ca compared to others (82.31 vs 62.05), but showed dramatic decline in phase II (30.66 vs 50.95, p<0.0001) and in phase IV (58.53vs 68.70, p<0.021). Single item symptoms showed significant declines related to fatigue, appetite loss, and pain in H&N Ca patients (p<0.0001) during RT course. Poor orodental hygiene in Indian head and neck cancer patients increases the incidence of post-irradiation periodontal disease and caries. Radiation to H&N region severely affects physical and social functions and global health status, significantly in mid-course and immediately after therapy.
24/ QUALITY OF LIFE AFTER ESOPHAGEAL RESECTION FOR CANCER AND BARRETT'S
Claude Deschamps, Francis C. Nichols, Mark S. Allen, Peter C. Pairolero, Daniel L. Miller, Victor F. Trastek, Division of General Thoracic Surgery, Mayo Clinic, Rochester, MN
Early detection and resection of esophageal carcinoma provide the best chance for cure. We reviewed and analyzed quality of life in 64 patients who survived more than 5 years after resection of esophageal carcinoma. We used the Medical Outcomes Study 36-Item Short-Form Health Survey (MOS SF-36). Physical function scores were decreased significantly (p<0.01) compared to the national norm. Ability to work, social interaction, daily activities, emotional dysfunction scores, and perception of health were similar to the national norm. Level of energy was decreased compared to the national norm but the significance was borderline (p = 0.05). Our patients had higher scores in the area of mental health (p<0.05). The occurrence of a postoperative anastomotic leak adversely affected the physical functioning and the health perception scores in our population (p < 0.05). Also, the need for postoperative dilatation adversely affected the social functioning score (p < 0.01). Age, sex, time interval since the surgery, location of lesion, histology, type of surgery, and adjuvant therapy did not significantly affect any of the 8 conceptual areas. More recently, 44 patients who had esophagectomy for Barretts with high-grade dysplasia (HGD) responded to the same written survey. Patients with HGD only (i.e. no cancer) find themselves to be better than the norm in role-physical and role-emotional (p<0.03). The occurrence of an anastomotic leak adversely affected the social functioning scores (p=0.02). Age was significantly correlated with physical functioning (r = -0.49, p = 0.0007) and role-physical (r = -0.33, p = 0.03). Time from surgery is significantly correlated with social functioning (r = 0.34, p = 0.02). While functional outcome after esophagectomy is acceptable but less than ideal, the operation had no consistent and measurable negative impact on these patient’s quality of life, as measured by the MOS SF-36.
25/ DIABETES AND CHANGE IN PSYCHOSOCIAL WELL-BEING
Siri Naess, John Eriksen, NOVA - Norwegian Social Research, Oslo, Norway
Previous research has documented that people with diabetes report lower quality of life than do people with no reported disease. In recent years new treatment regimes for diabetes have been introduced, including improved insulin, simpler blood sugar tests and transfer of responsibility from doctor to patient. Have these improved methods for controlling diabetes resulted in improved psychosocial well-being? In the this paper we analyse changes in psychosocial well-being between 1984-86 and 1995-97. On these two occasions the entire adult population of one county in Norway was called in for health screenings (the Nord-Trondelag Health Surveys, HUNT1 and HUNT2). The questionnaires included questions on several known diseases and impairments, self-assessed health and psychological well-being. 77,224 and 62,784 persons respectively responded to the questionnaires (response rates were 88.5 per cent in 1984-84 and 69.3 per cent in 1995-97). In a second questionnaire both in HUNT12 and HUNT2 more questions on psychosocial well-being were asked (with lower response rates, 72.6 and 58.0 per cent respectively). Preliminary analyses indicate that people with diabetes reported lower well-being than people with no reported diabetes, in HUNT1 as well as in HUNT2. Further, data indicate that the relationship between diabetes and well-being was weaker in HUNT2 than in HUNT1. More specifically, self-assessed health, the feeling of being strong and fit, the use of tranquillisers, and life satisfaction had improved during the 11 years between the two surveys, for people with diabetes compared to people with no reported diabetes. Reduced well-being was related to additional diseases (late complications), such as macrovascular disease. Controlled for reports on these complications, the differences between people with and without diabetes were slight. In the paper these changes in psychosocial well-being from 1984-86 to 1995-97 will be pursued in more detail by multivariate analysis techniques (like ANOVA and "sandwich"(Huber/White).
26/ THE REVISED KINDL-R: FINAL RESULTS ON RELIABILITY, VALIDITY AND RESPONSIVENESS OF A MODULAR HRQOL INSTRUMENT FOR CHILDREN AND ADOLESCENTS
Ulrike Ravens-Sieberer, Child and Adolescent Health, Robert Koch Institute, Berlin, Germany; Monika Bullinger, Department of Medical Psychology, University of Hamburg, Hamburg, Germany
Children’s self report of HrQoL has become an important outcome in pediatrics and public health. This paper reports the validation project of the 24 item generic self-report QoL questionnaire for healthy and ill children: the revised Kindl-R, intended to be used as a discriminant, predictive and evaluative instrument in community and clinical pediatric populations. In developing the Kindl-R, the goal was to design a modular approach to measure HrQoL in healthy and ill children with a generic core and disease specific modules as outlined by the WHO. The measure can be completed by children and parents as paper-pencil ond/or computer assisted version and is available for different age groups (4-7, 8-12, 13-16). Items were derived from focus groups with children. The measure (languages: German, English, French, Italian, Spanish, Greek, Swedish) can be augmented by disease-specific modules for obesity, bronchial asthma, atopic dermatitis, cancer and diabetes. 5781 healthy and chronically ill children and adolescents (4 to 17 years) and 3200 parents completed the Kindl-R in a number of longitudinal studies together with clinical and psychosocial predictors over a period of up to 2 years. Six dimension scores (physical, psychological well-being, family, friends, self-esteem, school) were computed as well as an index. Item response theory was used to determine optimal item and scale characteristics as well as multimethod-multitrait analysis and regression analysis. Content and construct validity was confirmed, factor validity was shown. Small to medium scale intercorrelation supported multidimensionality. The KINDL-R is reliable in terms of internal consistency (alpha >0.80) and test-retest results (r=0.69). Convergent validity was satisfactory. The acceptance of the measure was high(<15 min). The questionnaire has been able to distinguish between different acute and chronic health conditions (p<.05) and different health behavior(known groups). It has been proven to be sensitive to change over time in patients under treatment with moderate to high effect size estimates (Kazis) related to treatment (d > 0.6 in modules). The results support the KINDL-R as a flexible, modular, psychometrically sound measure for use in public health research and clinical applications.
27/ RETEST RELIABILITY AND VALIDITY OF A SURVEILLANCE MEASURE OF HEALTH-RELATED QUALITY OF LIFE
E. Andresen, T. Catlin, K. Wyrwich, Saint Louis University, St. Louis, MO; J. Jackson-Thompson, Missouri Department of Health, Jefferson City, MO
Measures of health-related quality of life (HRQoL) are increasingly common to clinical research, but uncommonly used in surveillance systems. For surveillance use, HRQoL measures need to be brief, generic, and broadly applicable to public health questions. HRQoL has been proposed as a method to monitor health in the American public health plan, "Healthy People 2010." We investigated the reliability and validity of the 4 HRQoL surveillance questions from the Behavioral Risk Factor Surveillance System - an ongoing nation-wide random-digit dialed telephone survey of Americans (aged ³18). In 1999, randomly sampled adults from a midwestern state were re-interviewed for a retest of the HRQoL questions; 868 adults completed both interviews. The HRQoL questions include Self-Reported Health and 3 "days" questions measuring poor physical health, poor mental health, and activity limitation in the past 30 days. Poor mental and physical health days combine for a positive summary measure of Healthy Days. Mental and physical health days were also dichotomized using a cutoff of 14 days. Reliability is reported as kappa statistics for categorical questions and intraclass correlation coefficients for continuous questions. Validity of continuous "days" questions was examined by linear trend analysis across Self-Reported Health categories. Retest reliability was excellent (0.75 or higher) for Self-Reported Health and Healthy Days measures, and moderate (0.58 to 0.71) for other measures. Reliability was lower for older adults. Other demographic subgroups (e.g., gender, race/ethnic group) showed no regular pattern of differing reliability. Validity of the "days" questions showed strong linear trends across categories of Self-Reported Health, especially for physical health days. Retest reliability of the HRQoL surveillance measures is moderate to excellent, and internal validity is strong. Scaling options will require future attention, as will research into appropriate metrics for what constitutes important population differences.
28/ THE SEXUAL RELATIONS SCALE: PSYCHOMETRIC CHARACTERISTICS IN PATIENTS WITH HIV DISEASE, GERD, UTERINE FIBROIDS, AN PERIPHERAL ARTERIAL DISEASE
Dennis A. Revicki, Karin Coyne, Carol Koro, Center for Health Outcomes Research, MEDTAP International, Bethesda, MD; Albert W. Wu, Health Services Research and Development Center, Johns Hopkins University, Baltimore, MD; Anne Rentz, Center for Health Outcomes Research, MEDTAP International, Bethesda, MD
Many chronic diseases and medical treatments impact on patient sexual functioning and behavior. Impairments in sexual functioning may decrease health-related quality of life, especially social and intimate partner relationships. Brief, psychometrically sound instrument are needed to evaluate sexual functioning and relations for clinical trials and clinical monitoring. The objective of this study was to evaluate the reliability and validity of the Sexual Relations Scale (SRS) in patients with HIV disease (n=163), GERD (n=138), uterine fibroids (n=139), or peripheral arterial disease (PAD) (n=60). All patients were recruited from clinical centers and the sample was 57% women and 50% non-white, with average ages of 31 to 63 years, depending on the sample. The SRS consists of 6 questions evaluating intimate relationships: health interference with sexual relations or ability, satisfaction with sexual relations, decrease in sexual activity, and intimate relations with partner. For all four samples, the SF-36 Health Survey and disease-specific HRQL measures were included; for the HIV disease and uterine fibroid samples, the MOS Sexual Problems Scale was included; and for the HIV sample, the CESD was also included. Internal consistency reliability (Cronbach's alpha) ranged from 0.84 to 0.92 across patient samples. The SRS was significantly correlated with the MOS Sexual Problems Scale in men (r=0.82, p<.05) and women (r=.73, p<.05) with HIV disease and in uterine fibroid patients (r=.71, p<.05). Patients with AIDS or symptomatic HIV disease reported more impairments on SRS scores compared with those who had asymptomatic HIV disease (p<.05). SRS scores were most strongly correlated with SF-36 social function, general health perceptions, vitality and mental health scores. In the HIV sample, SRS scores varied significantly by depression status (p<.0001). Uterine fibroid patients reported mean SRS scores that were significantly more impaired than controls (p<.05), and SRS scores were significantly correlated with a disease-specific HRQL measure (r=.34, p<.01). Based on this study, the SRS had acceptable reliability and evidence of construct validity across several chronic disease populations. Future research needs to evaluate the responsiveness of the SRS and determine guidelines for interpreting clinical meaningfulness of SRS scores.
29/ VALIDATION OF THE UFS-QOL, A DISEASE-SPECIFIC SYMPTOM AND HEALTH-RELATED QUALITY OF LIFE QUESTIONNAIRE FOR UTERINE FIBROIDS
Karin Coyne, Center for Health Outcomes Research, MEDTAP International, Bethesda, MD; James Spies, Department of Inter-Radiology, Georgetown University, Washington, DC; Deneane Boyle, Center for Health Outcomes Research, MEDTAP International, Bethesda, MD; Kerry Skyrnarz-Murphy, Sheila Gonzalves, Department of Inter Radiology, Georgetown University, Washington, DC
Uterine fibroids (UF) cause significant symptoms and health-related quality of life (HRQL) impairment in women who experience them. The purpose of this research was to validate a newly created instrument to assess symptom severity and HRQL in women with UF. The UFS-QOL was generated from focus groups, literature reviews and clinical opinion. Participants (pts) for the validation study were recruited from 5 area gynecologist offices, an interventional radiology department, and a University campus. Instruments used for validation were the SF-36, Ruta’s Menorrhagia questionnaire, the Revicki-Wu sexual functioning scale, and physician and patient assessments of severity. Item and exploratory factor analysis were performed to assess the subscale structure of the questionnaire. Psychometric evaluation was conducted to assess reliability and validity. Test-retest was performed on a random subset of the study group within 2 weeks of the initial visit. 110 patients with confirmed fibroids and 30 normal subjects participated in the validation. Using defined decision rules for item reduction, the questionnaire was reduced to 8 symptom questions and 29 HRQL questions with 6 subscales. Subscale cronbach’s alphas ranged from 0.81 to 0.94. Small to moderate correlations were present with the SF-36 subscales, Menorrhagia questionnaire and the Revicki-Wu sexual functioning scale. The UFS-QOL showed excellent discrimination among UF patients with varying degrees of symptom severity and from normals. Subscale intra-class correlations for two week retest reliablilty ranged from 0.76-0.92. The UFS-QOL appears to be a useful new tool for detecting differences in symptom severity and HRQL among patients with uterine leiomyomata. Further investigation regarding the responsiveness of the UFS-QOL to therapies for UF is underway.
30/ PSYCHOMETRIC EVALUATION OF TWO OBESITY AND WEIGHT LOSS QUALITY OF LIFE INSTRUMENTS: THE OWLQOL AND WRSM
Donald L. Patrick, Department of Health Services, University of Washington, Seattle, WA; Donald M. Bushnell, Health Research Associates, Seattle, WA; Laura A. Glauda, Dennis D. Gagnon, Margaret Rothman, The R. W. Johnson Pharmaceutical Research Institute, Raritan, NJ
To report U.S. validation of two measures specific to obesity and weight loss: the Obesity and Weight-Loss Quality of Life Instrument (OWLQOL) and Weight-Related Symptom Measure (WRSM), including measurement model, psychometric properties and burden. Two groups were recruited: 160 obese enrollees in diet, exercise, and health programs, and 180 obese non-enrollees from the general public. Respondents had a Body Mass Index (BMI) of 27 through 50 and were between the ages of 18-75. All completed the OWLQOL, the WRSM, the SF-36, the CES-D (depression), the Perceived QOL scale, and standard demographic items at baseline, 1 week (for test/retest of OWLQOL and WRSM), and again at 12 weeks (for responsiveness based on weight loss and global rating of change). Psychometric testing followed standardized procedures. 340 patients completed baseline assessments: 60.0% female, 77.9% white, and 50.3 married. The overall OWLQOL score was internally consistent (alpha=0.96) and reproducible (ICC=0.90). As predicted, OWLQOL scores had stronger associations with the SF-36 vitality subscale (0.54), the more generic quality of life scale (QCOL, 0.54), weight-related symptom bothersomeness (0.53), and depressive symptomotology (CES-D, 0.49) than the SF-36 physical function subscale (0.40). The overall WRSM was also internally consistent (0.87) and reproducible (0.83). OWLQOL scores were able to discriminate between levels of BMI, levels of symptoms, levels of depressive symptomotology (CES-D), and gender. The total sample (irregardless of being in a program) was evaluated for responsiveness at 12 weeks. Effect size for 2.5% or greater decrease in the patients’ weight, were 0.76 for the OWLQOL, 0.54 for the WRSM, and ranged from 0.20-0.60 for the 8 subscales of the SF-36. In this stand alone study, the OWLQOL and WRSM proved valid, reproducible, and responsive to weight-loss and global rating of change.
31/ COPING STYLES PREDICT QUALITY OF LIFE IN ADOLESCENTS WITH INFLAMMATORY BOWEL DISEASE
Hester J. Loonen, Bert H. Derkx, Pediatrics, Martha A. Grootenhuis, Bob F. Last, Psychosocial Department, Emma Children's Hospital, Academic Medical Centre, Amsterdam, The Netherlands
In clinical practice, more attention is being paid to ways to ameliorate patients' QoL. Inflammatory bowel disease (IBD) is a chronic relapsing disorder with heavy treatment protocols that challenges patients' coping abilities considerably. The aim of this study was to analyse which variables (including sex, disease activity and reliance on certain coping strategies) best predict HRQoL. We used a cognitive control strategy scale (CCSS), which measures four coping styles: predictive coping (having positive expectations about the future), vicarious coping (attributing power to medical caregivers and treatment), interpretative (searching for meaning and information), and illusory control (wishful thinking), and we used a disease-specific HRQoL questionnaire (Impact-II (NL)), measuring six domains of QoL (bowel and systemic symptoms, emotional and social functioning, body image and treatment related concerns). Adolescents with IBD (12 to 18 years old) were asked to fill out the CCSS and the Impact-II (NL). To investigate which variables predict adolescents HRQoL, regressions analyses in steps were performed with the Impact domains as the dependent variable, and sex, diseease activity and predictive and vicarious control as predictor variables. More reliance upon predictive (optimism about course of disease) and vicarious coping (attributing power to medical care givers) predicts better QoL. Gender and disease activity also attributed to the model, although less than the coping styles. Coping is an important predictor of HRQoL, even more than the widely accepted gender and disease activity variables. Our findings demonstrate that interventions with adolescents with IBD could focus on enhancing the efforts to appraise their disease more positively, and on strengthening the reliance upon the doctor, besides traditional therapies aimed at decreasing disease symptoms.
32/ "I THINK MY LIFE IS BETTER THAN MOST KIDS" LIFE SATISFACTION IN HEALTHY AND CHRONICALLY ILL CHILDREN
Kirsten Moenkemoeller, Kinderkrankenhaus, Universität Witten/Herdecke, Köln, Germany; Anne Wrede, IFAS, Universität zu Köln, Köln, Germany; Guido Buerk, Vestische Kinderklinik, Universität Witten/Herdecke, Datteln, Germany; Ulrike Ravens-Sieberer, Forschungsgruppe, Robert-Koch-Institut, Berlin, Germany
Life satisfaction (LS) as the cognitive component of well being refers to a "judgemental process, in which individuals assess the quality of their lives on the basis of their own unique citeria" (Shin, D.C., Johnson, D.M.1978). Do chronically ill children judge their quality of life any different than healthy children when applying their own criteria? The "Student's Life Satisfaction Scale" for children aged 7-14 (Huebner, E. 1991) was translated into German and adapted for children aged 8-12. Additionally the satisfaction in five domains (school, family, friends, living environment, and self) was registered applying an adapted version of the "MSLSS" (Huebner, E. 1994). The questionnaires were administered to a group of 347 children aged 8-12. 130 children with Asthma (n=39), Diabetes (n=45) and Rheumatic Diseases (n=46) were interviewed using the revised version. Teachers of healthy children and parents of the chronically ill were interviewed as external raters. Both scales showed acceptable psychometric properties. Convergent validity could be shown with teachers and parents ratings in expected directions. T-Test showed a significant difference between the global LS of the two groups (p=.001). Applying Linear Regression, the satisfaction in the five domains explained 47% of global LS for healthy children but only 27% for the chronically ill. Including all variables in a Backwards Linear Regression, sickness parameters explained 9% with function and severity rated by the children being the strongest predictors. Logistic Regression showed that 86% of the extremely satisfied/unsatisfied children (SD>1) can be classified by the MSLSS and 87% applying all variables. Evaluating specific living domains as well as disease specific concepts might not grasp the criteria the children apply to judge their quality of life. Complementing HRQL interviews by global LS ratings can enhance the potential to classify healthy or chronically ill children who need special support.
33/ QOL AFTER PEDIATRIC STEM CELL TRANSPLANTATION: A CASE FOR RESPONSE SHIFT?
Susan K. Parsons, Christopher J. Recklitis, Dana-Farber Cancer Institute, Boston, MA
Hematopoietic stem cell transplantation (HSCT) raises central quality of life (QOL) issues and significant ‘adaptational’ challenges for survivors. Treatment is intense, including risk of sequelae and fear of recurrence. In pediatric HSCT, QOL evaluation is complicated by reliance on parents’ proxy report. The transplant affects parents and their adaptation has a major impact on the child. To understand children's QOL after HSCT both the children and parents’ ratings and the interaction between the ratings over time must be considered. We performed a cross-sectional analysis using the Child Health Ratings Inventory and the Disease-Specific Impairment Inventory-HSCT of 150 children (5-12 yrs.) and adolescents (13-21 yrs.) after HSCT. For children surveyed soon after transplant (i.e., < 6 mo.), parent reports were universally lower in all dimensions (~1-2 s.d.; p<0.001). Later post transplant (i.e., > 12 mo.), a dramatic shift occurred with parent scores equaling or exceeding all child scores. These parent-child ratings were not correlated in either time period. Child self-reports,however, were more strongly correlated with disease severity than parent reports. The pattern in the adolescent sample is slightly different. Early on, parents scored significantly lower than adolescents in only two of the seven dimensions. Later on, no significant differences were detected between raters, and correlations within parent-adolescent pairs were robust (r=0.36-0.97). In contrast to the younger sample, adolescent self-report was generally not associated with disease severity, whereas their parents’ratings were. These results suggest that response shifts may differ as a function of age and cognitive development. Further, over time the parent and the child derive different meaning from the same external reality. Parental scores may reflect an internal recalibration, influenced by the child’s survival and resumption of normative activities. The data will be discussed within the context of a response shift model.
34/ PEDIATRIC SPINE DEFORMITY: ASSESSING PATIENT OUTCOMES AND QUALITY OF LIFE
Michael G. Vitale, Pediatric Orthopaedics, Children's Hospital of New York, New York, NY; Julie C. Choe, Douglas E. Levy, Pediatric Orthopaedics, Annetine C. Gelijns, Alan J. Moskowitz, Surgery, InCHOIR, New York, NY; Joshua E. Hyman, Francis Y. Lee, David P. Roye, Pediatric Orthopaedics, Children's Hospital of New York, New York, NY
Several new outcome measures have been recently published to assess quality of life (QOL) outcomes in pediatric orthopaedics and pediatrics, in general. Although these measures have shown promise, none have been validated for use in children with scoliosis. Parents of 279 children with scoliosis completed the Child Health Questionnaire (CHQ) and the American Academy of Orthopaedic Surgeons (AAOS) Pediatric Outcomes Instrument between 1997 and 2000; the Scoliosis Research Society (SRS) instrument was completed directly by the affected children. The treating physician completed a subjective rating of health and documented pertinent socioclinical data for each child. Scores were compared with one another and to the clinical parameters. The effect of surgery on QOL (responsiveness) was assessed for 40 patients who had scoliosis surgery. Ceiling effects were noted for each instrument, especially in the physical function and self esteem domains. However, patients with scoliosis scored significantly lower than published values for "healthy" children in several CHQ domains (t-test): physical function (p<0.005), bodily pain (p=0.01), parental time impact (p=0.02), parental emotional impact (p<0.005), behavioral scale (p<0.005). Moreover, patients with larger curves exhibited lower QOL scores, as measured by the CHQ, supporting the face validity of this measure. In general, the CHQ showed superior psychometric characteristics over the AAOS instrument in this population. Four SRS domains were correlated with degree of curvature: physical function (r=-0.17, p=0.008), bodily pain (r=-0.32, p<0.005), self image (r=-0.24, p<0.005), and satisfaction (r=-0.29, p=0.02). Preliminary analysis of longitudinal data on patients at 1-year after surgery showed no significant changes. The CHQ, AAOS, and SRS instruments all suggest that scoliosis has a negative effect on the physical and psychosocial health status of affected children.
35/ HEALTH RELATED QUALITY OF LIFE OF CHILDREN WITH CHRONIC DISEASES
Martha A. Grootenhuis, Bob F. Last, Psychosocial Department, Emma Children's Hospital, Academic Medical Centre, Amsterdam, The Netherlands
Advances in pediatrics and pediatric surgery, have led to the need to evaluate the physical, psychological and social consequences of medical treatment in children. In several patient populations Health Related Quality of Life (HRQOL) has been measured with the TACQOL-CF, which measures the domains: physical, motor, cognitive and social functioning, autonomy, and positive and negative emotions. Chronically ill children aged 8-18 years with different diagnosis: inflammatory bowel disease (n=82), congenital heart disease (n=73), familial hypercholesterolemia (n=69), end stage renal failure (n=18) and cancer (n=17) participated in different ongoing studies. Aim of this study was to compare the HRQOL of children with a chronic illness with healthy controls and to investigate within-group differences. Children with a chronic disease were compared with a random sample of 400 healthy children aged 8-15 years using t-tests. Analysis of variance were used to investigate differences between illness-groups. Children with a chronic illness (complete sample n=259) differed from healthy children on the domains of motor functioning and autonomy with lower HRQOL. Analysis of variance showed differences between illness groups on the domains of motor and cognitive functioning and on autonomy. Major findings are: Children with a congenital heart disease reported lowest HRQOL on motor functioning (difficulties with walking, running and power of endurance). Children with cancer report lower quality of life in autonomy (doing things on their own or independently) whereas children with end stage renal failure report affected HRQOL on cognitive functioning (concentration, difficulties with schooltasks). Although all specific groups have their own issues, the findings show some of the disease-specific problems of children with a chronic illness. Intervention programs should take these problems into account.
36/ THE RELATIONSHIP BETWEEN RESOURCE CONSUMPTION AND ASTHMA SEVERITY AMONG ASTHMATIC OUTPATIENTS IN THREE EUROPEAN COUNTRIES
David U. Himmelberger, Health Outcomes Group, Palo Alto, CA; Leonardo Antonicelli, Hospital Umberto I, Ancona, Italy
This study was designed to evaluate the relationship between asthma severity as measured by the GINA Guidelines (intermittent, mild, moderate and severe persistent) and resource consumption among patients with asthma. A total of 1,128 patients from Spain, Italy and France participated in the study. Data were collected from patients based upon their self reported recall using a questionnaire designed to measure the relevant components of resource consumption and asthma severity. Additional data about the patients’ asthma status and severity was collected from their physicians. The data of this study confirm the hypothesis that more resources are consumed as the severity of asthma increases. The total resource consumption in monetary units varied from 6-14.5 times greater for patients with severe persistent asthma than for patients with intermittent asthma ($2,720 vs. $428 for Spain and $19,670 vs. $1,350 for Italy).
As severity increased from intermittent to severe persistent, a 2-3 fold increase in limitation in daily activities was observed. Significant increases in time lost from paid and non-paid work were also observed. Peak flow was below the predicted values for all four GINA classifications of severity, with the magnitude of the differential varying greatly across the three mildest GINA classes (8-29%) and was slightly greater for the patients with severe persistent asthma (32-45%).
Similar to the results with limitation of daily activities and asthma symptoms, consumption of individual resources increased two to three fold in terms of natural health units in all areas measured as the severity increased from intermittent to severe persistent asthma.
The data from this study suggests that resource consumption is more strongly related to asthma severity, as measured by the GINA guidelines, than it is to pulmonary function, as measured by peak flow.
37/ THE EFFECT OF ASTHMA SEVERITY ON HEALTH-RELATED QUALITY OF LIFE AND WORK PERFORMANCE: DOES THE DEFINITION OF SEVERITY MAKE A DIFFERENCE?
Steven R. Erickson, Duane M. Kirking, College of Pharmacy, University of Michigan, Ann Arbor, MI; Nancy Kline Leidy, Center for Health Outcomes Research, MEDTAP International, Bethesda, MD
This study compares HQL and WP scores across five methods of determining asthma severity: three subjective methods: patient perceived severity (PPS); patient-reported symptom-severity-overall (OSS); symptom-severity nocturnal (NSS); and two objective methods using retrospective claims analyses: oral steroid use (OSU) and multi-drug use (B2, inhaled and oral steroids). Three levels of severity were examined: mild, moderate and severe. Secondary analysis of data from a cross-sectional mail survey of 603 adults treated for asthma in a U.S. managed care organization linked to pharmaceutical/medical claims data. Patients completed the SF-36, Asthma Quality of Life Questionnaire (AQLQ) and a WP scale. Analysis of variance (ANOVA) procedures were used to compare scores across severity level within method and across method within severity level. Within each method, significant severity effects were found for PCS, AQLQ and WPS (except OSU). Significant effects for MCS were also found in the OSS and NSS models. In post-hoc analyses, mild-moderate and mild-severe comparisons were consistently significant (p<0.05). In the cross-method analyses, significant method effects were found at all levels of severity (p<0.05) for most scale scores with ranges as follows: Mild-PCS 47.7 to 50.0; MCS 48.7 to 49.8; AQLQ 5.3 to 5.8; WPS 89.5 to 93.3. Moderate-PCS 41.7 to 47.5; MCS 46.4 to 48.7; AQLQ 4.0 to 4.8; WPS 81.9 to 88.7. Severe-PCS 38.3 to 43.6; MCS 44.1 to 49.2; AQLQ 3.8 to 4.7; WPS: 75.4 to 85.7. Within-method results were consistent with previous studies demonstrating a significant relationship between severity, HRQL and WP. HRQL and WP scale scores varied by the method used to define severity, however. This variation should be considered when performing cross-study comparisons of burden of illness and treatment effects.
38/ QUALITY OF LIFE AFTER INTENSIVE CARE FOR ACUTE LUNG INJURY
Cynthia R. Gross, Academic Health Center, University of Minnesota (UMN), Minneapolis, MN; Craig R. Weinert, Medical School, Minneapolis, MN; Carol Albright, Caron L. Bury, William A. Marinelli, UMN, Minneapolis, MN
Progress has been made in reducing in-hospital mortality from Acute Respiratory Distress Syndrome (ARDS), but there is still relatively little information on the health status of survivors. A sample of 51 patients (men=29, women=22, aged 19-78) who recovered from ARDS or acute lung injury completed a battery of mailed, self-report questionnaires at six months or longer after discharge from intensive care (median 8 mos). Their health-related quality of life (HRQL), as measured by the SF-36, showed large and statistically significant decrements compared to norms across all domains. Clinically relevant levels of depression symptoms were found in 46% of the sample based on the Center for Epidemiologic Studies Depression Scale (CES-D). A subsample (N=18) completed the Post-Traumatic Stress Disorders Checklist (PCL), and while only one patient scored above the level considered indicative of the diagnosis, 61% (11/18) reported these symptoms interfered with their normal social activities. In our sample, the SF-36 mental component score is strongly correlated with the PCL (r=-.80, N=18, p<.01) and with the CES-D (r=-.83, n=49, p<.01). Depression and PCL scores were positively correlated (r=.64, p<.01). These findings highlight the need to include HRQL outcomes in ICU-based clinical trials. Currently, our group is investigating the relationships among factors such as medications (sedatives and neuromuscular blocking agents) or therapies (ventilation strategies, new medications) and these HRQL outcomes. Additional research should establish whether clinicians are aware of these long-term decrements in HRQOL or how families and patients are prepared for post-ICU sequelae. Funding: SCOR Clinical Core P50 HL50152.
39/ EFFECTS OF DISEASE MANAGEMENT ON QUALITY OF LIFE IN PATIENTS WITH ASTHMA
Wolfgang Greiner, Matthias J. Graf v.d. Schulenburg, Institut für Versicherungsbetriebslehre, University of Hannover, Hannover, Germany
In the last years, Disease Management has been recognized as a significant element in improving the care situation of chronically ill people in particular, both from a medical and an economic point of view. This study deals with an economic evaluation of a Disease Management Programme for asthma with the help of an electronic control instrument for Home-Monitoring of the patient. Research was made on the improvement in the quality of life as well as the saved costs. 112 adult patients with asthma of varying severities were divided randomly into two groups. The patients are all members of a large German health insurance group in the Bundesland Rheinland-Pfalz. The first group received an extensive Disease Management Programme with electronic Home-Monitoring, while the control group was treated further by their GP. The patients were observed over a period of 12 months. Lung function tests and investigations on the quality of life were carried out in intervals of six months (at the beginning, after six months and after 12 months; visits one, two and three). The quality of life was measured using the questionnaires Quality of Life Asthma (FLA), a validated disease-specific questionnaire and the EQ-5D (Euro-Qol). In addition, the treatment costs during the study time period were documented and evaluated. During the study, the quality of life of the control group revealed no changes statistically, while the patients in the Disease Management group achieved significantly higher values than those at the beginning of the study. The improvements are related particularly to the "Physical and Psychic Attributes" of the FLA, although they can also be seen in the (less sensitive) generic EQ-5D. Thus, the average EQ-5D value of 62 increased to 72, while in the control group it dropped from 65 to 60. The study data additionally shows that considerable savings are possible with Disease Management for asthma, especially concerning hospital services. These savings are partly counteracted with higher costs in other area.
40/ COPING AND QUALITY OF LIFE IN ADULTS WITH ASTHMA
Edgar A. Correia, Instituto Superior de Psicologia Aplicada, Lisboa, Portugal; Jose L. Pais-Ribeiro, Faculdade de Psicologia, Universidade do Porto, Porto, Portugal; Nuno B. Neuparth, Pedro L. Mata, Fisiopatologia, Faculdade de Ciencias Medicas, Universidade Nova, Lisboa, Portugal
The objective of this cross-sectional study is to explore the relationships between coping strategies and health related quality of life (HRQOL) in patients with asthma. The portuguese versions of the Coping with Health Injuries and Problems Scale (CHIPS) and the Asthma Quality of Life Questionnaire (AQLQ) were applied to 60 adults, from both sexes (52% females), with asthma diagnostic. Results showed some significant correlations between self-reported coping strategies and HRQOL in asthma. Patients who scored higher in Palliative Coping reported significant worst HRQOL in all domains of the AQLQ (varying from r (60) = -.27, p<.05 in the Emotional Function domain, to r (59) = -.39, p<.01 in the Exposure to Environmental Stimuli domain), as well as in the Overall Quality of Life, r (58) = -.35, p<.01. Similar pattern appeared when correlation’s between Emotional Preoccupation coping and AQLQ were maid: effectively, asthmatic patients that reported more use of Emotion-Oriented strategies to deal with their asthma showed significant lower HRQOL (p<.01, in the overall AQLQ, as well as in all its domains). Otherwise, patients that referred more resource of Instrumental or Distraction coping had only negative correlations with Environmental Exposure related QOL (r (59) = -.32, p<.05; r (59) = -.34, p<.01, respectively). Results suggests, having in account that some ways of coping have clearly negative implications in HRQOL in patients with asthma, that it can have advantage in creating self-management education programs that attend to psychological factors, namely coping strategies.
41/ USING ITEM RESPONSE THEORY TO IMPROVE QOL MEASURES FOR CHILDREN — FIRST RESULTS FROM THE CHILD DYNAMIC HEALTH ASSESSMENT PROJECT
Ute Ellert, Christiane Thomas, Ulrike Ravens-Sieberer, Department for Child and Adolescent Health, Robert Koch Institute, Berlin, Germany; Mark Kosinski, Quality Metric Inc., Lincoln, RI; Jakob B. Bjorner, National Institute of Occupational Health, Copenhagen, Denmark, Jim Dewey, John E. Ware, Quality Metric Inc., Lincoln, RI
The aim of the project was to apply Item Response Theory Models to equate and score 3 widely used health related quality of life questionnaires for measuring the emotional well-being of children. Data from a rehabilitation study (N=881) and a survey with healthy children (N=185) was analyzed. The children and adolescents (8-16 years) completed three different HrQoL instruments (KINDL-R, CHQ, SF-36) pre-treatment and directly, 3 months, and 12 months post-treatment. The total pool of items measuring mental health was 28. Factor analytic methods were used to test the assumptions of IRT, unidimensionality and local independence if items. IRT methods were used to evaluate the ordinal consistency of response choices and the distinctiveness of items representing a "mental health" dimension for all instruments and to score an overall child mental health concept (theta). Measurement precision using IRT based scoring was evaluated against the developer’s scoring for each instrument and all were evaluated in relation to theta. Eight items were dropped since they did not meet the IRT assumptions of unidimensionality or local independence. Response choices to more than one third of the mental health items showed ordinal inconsistencies. Model fit was improved by re-scoring the response choices of these items. For the IRT-based scale scores statistically significant gains in measurement precision were observed compared to the scores based on the developer’s method. The pool of items from all instruments combined provided better coverage of the range of subjective mental health measured (theta) in comparison with mental health domains covered by each instrument. The item pool was included in a ChildDynHA demo version running on the Internet and its feasibility was tested on 60 school-aged children indicating good acceptance and a minimum of time and effort for the children. In conclusion, IRT methods are likely to be useful in improving measurement precision and comparing results across different instruments that measure the same health concepts in children. Computer adaptive administration of items from a pool of items from all instruments is likely to yield more precise measures of child health and reduce respondent burden.
42/ ITEM BANKING TO IMPROVE, SHORTEN AND COMPUTERIZE SELF-REPORTED FATIGUE
Jin-shei Lai, David Cella, Chih-Hung Chang, Evanston Northwestern Healthcare (ENH), Northwestern University (NU), Evanston, IL; Rita K. Bode, Rehabilitation Institute of Chicago (RIC), Allen W. Heinemann, RIC, NU, Chicago, IL
Fatigue is a common symptom among cancer patients and the general population. Modern computerized adaptive testing (CAT) can enable precise assessment of fatigue across diverse groups using a small number of items from a fatigue item bank. This article illustrates 4 steps to prepare such an item bank, using 13 items from the Functional Assessment of Chronic Illness Therapy-Fatigue Subscale (FACIT-F) as the basis. Samples included 1,022 cancer patients and 1,010 people from the United States general population. Steps included: (1) select core items, (2) partition the fatigue continuum into probabilistic segments using a pre-selected criterion item, (3) position item response categories of the selected core item onto the fatigue continuum, and (4) create an item bank matrix for CAT application. Using an Item Response Theory-based measurement model, Rasch measurement model, we selected 9 items based on their fit to the underlying model and their lack of differential item functioning. The items were then positioned on the fatigue continuum. The fatigue level for 66.8% of the general population and 82.6% of the cancer patients was well-captured by these 9 items. A precise fatigue score was obtainable using as few as 4 of the 10 banked items (including 1 criterion item from the FACIT-general). This study demonstrates that 10 banked items can precisely measure fatigue, but also that a more extensive item bank covering the entire fatigue continuum will further enhance measurement precision. Four steps to create the item bank and derive the CAT algorithm will be presented.
43/ ONE- VERSUS TWO-PARAMETER ITEM RESPONSE THEORY (IRT) MEASUREMENT MODELS APPLIED TO MOS PF-10 SCORES: HOW MUCH DOES IT REALLY MATTER?
Chih-Hung Chang, David Cella, Center on Outcomes, Research and Education, Evanston Northwestern Healthcare, Evanston, IL
Item response theory (IRT)-based models have increasingly been applied in health status questionnaire development and assessment. There is debate as to which of these models, often classified as 1- versus 2-parameter (P) models, is most appropriate. This study examined the comparability of original and IRT scoring methods for the MOS 10-item Physical Functioning (PF-10) scale. The PF-10 data were collected from 5 sites as part of a larger project to equate five commonly used quality of life instruments in a total of 1,714 cancer and HIV/AIDS patients. Physical functioning scores were obtained using algorithms in Ware et al (1994) and Hays et al (1998), as well as two 1-P models: Masters’ partial credit model (PCM); Andrich’s rating scale model (RSM), and two 2-P models: Samejima’s graded response model (GRM) and Muraki’s generalized partial credit model (GPCM). Scores obtained by Hays, PCM, and RSM methods were all highly correlated with the original PF-10 scoring as in the SF-36 manual (all r > .975, p < .0001), while scores based on GRM and GPCM correlated only moderately with the original scoring (r=.60, p =.02). The two possible misfitting items (PF01 and PF10) identified by RSM and PCM also had the lowest slope parameters obtained from the GRM and GPCM, evidenced by the high degree of association between fit statistics in the RSM and PCM versus slope parameters in the GRM and GPCM (r =-.76 to -.77). Item difficulty (location) hierarchies obtained from the four IRT models were almost identical (correlation range =.998 to .999), indicating these models yield almost identical ordering of patients in measuring physical functioning. We conclude that these approaches yield comparable results and have good potential for providing complementary rather than conflicting data.
44/ DO MEASURES OF FUNCTIONAL ABILITY FULFIL THE ASSUMPTIONS ON DIMENSIONALITY REQUIRED BY ITEM RESPONSE THEORY?
Rebecca Holman, Robert Lindeboom, M. Vermeulen, Rob de Haan, Clinical Edpidemiology and Biostatistics, Academic Medical Center (AMC), Amsterdam, The Netherlands
One of the most important determinants of the level of health related quality of life enjoyed by patients is their functional health status. This is often defined in terms of their ability to perform activities of daily life and measured using multi-item instruments. The dimensionality of these instruments has often been well considered using classical psychometric methods, such as factor analysis. However, recently there has been a surge in interest in the application of item response theory to the analysis of such multi-item instruments. Unidimensional item response theory models are now relatively well known, and their utility in modelling and analysing the responses to multi-item instruments has been proven, in many clinical fields, including the measurement of HRQOL. However, interest is beginning to move towards models, which can accommodate a number of measurement dimensions. In this article we will examine a multidimensional item response theory model and how it can be used to shed light on the dimensionality of the instruments developed to measure the functional status. The item bank developed during the Amsterdam Linear Disability Score project to measure the functional status of chronically ill, but stable patients, will be considered in depth. The item bank consists of approximately 200 items and was constructed using item response theory techniques with the aim of placing all items on a single measurement scale. We will present the results of both multi- and unidimensional analyses of the data and consider whether the results are worth the extra input required to estimate a multidimensional item response theory model. We hope that the results obtained in this paper will provide some important guidelines for considering the dimensionality of multi-item instruments for the measurement of functional status when using item response theory techniques.
45/ FACTORS ASSOCIATED WITH STABILITY IN QUALITY OF LIFE RATINGS AMONG PEOPLE LIVING WITH HIV/AIDS: RESPONSE SHIFT’S "SMOKING GUN"?
Bruce D. Rapkin, Rosy Chhabra, Carolyn Springer, Psychiatry and Behavioral Sciences, Memorial Sloan-Kettering Cancer Center, New York, NY; Bruce Agins, Clemens M. Steinbock, AIDS Institute, New York State Department of Health, New York, NY; Mark Sharp, Ira Feldman, AIDS Institute, New York State Department of Health, Albany, NY
During intervals when global QOL ratings remain constant, significant changes may nevertheless occur in component quality of life scales. Such underlying changes necessarily indicate response shifts in the appraisal of overall QOL. In this paper, we will examine this phenomenon in order to better to understand how people living with HIV/AIDS maintain their overall QOL. Data are based on 1307 Medicaid recipients interviewed at two or more points in time. Measures included the MOS-HIV, health status, service use, life events, and demographics. Initial analyses compared change from first to last interview, average of 3 years (range 6 to 48 months). Of 1307 observations, 460(35.2%) provided the same answer to the overall health rating at first and last interviews. However, these patients displayed variance and covariance among MOS subscales similar to the full sample. Cluster analysis based on the stable subsample revealed six patterns of change in MOS subscales. 43.2% demonstrated stability on all MOS subscales. 56.8% showed change on one or more subscales despite stability in overall health. Discriminant analysis suggested three major patterns of sub-scale change: (1) Increases in emotional well being and energy offsetting losses in cognitive, social and role functioning; (2) General improvement in social, cognitive and role functioning; and (3) Improved emotional well-being and cognitive functioning offsetting decline in physical functioning and increased pain. Changes were associated with longer intervals between interviews, role changes, substance use and declining health. Findings demonstrate processes involved as individuals adapt to changing circumstances while maintaining their sense of well-being. Implications of these results for understanding how people modify their perspectives to maintain QOL will be discussed.
46/ RESPONSE SHIFT EFFECTS OF AN ADVANCE CARE PLANNING INTERVENTION WITH AMBULATORY GERIATRIC PATIENTS
Carolyn E. Schwartz, Department of Family Medicine & Community Health, University of Massachusetts Medical School, Worcester, MA; Noreen Basque, VNA Care Network, Worcester, MA; George Reed, Yungsheng Ma, Medicine, H. Brownell Wheeler, Surgery, University of Massachusetts Medical School, Worcester, MA
There is a large discrepancy between the wishes of dying patients and their actual end-of-life (EOL) care, with a negative impact on patients and survivors alike. It is postulated, however, that early intervention to facilitate advance care planning (ACP) can reduce this discrepancy and may lead to response shifts in quality of life (QOL). This randomized trial (n=61) evaluated the impact of the LaCrosse early intervention in planning EOL care for ambulatory geriatric patients, and investigated how patient's values, conceptualizations, and internal standards of QOL change as they engage in ACP. Patients were randomized either to a non-directive control condition or to an intervention in which an ACP-trained nurse facilitated a discussion between the patient and his/her health care proxy. They discussed the benefits and burdens of life-sustaining treatments, and the patient’s goals and preferences for life-sustaining treatments. Two-month follow-up revealed that the intervention achieved higher congruence between patients and proxies in reporting patients’ EOL care preferences with 81% in complete agreement vs 57% [Effect Size (ES) of difference in overall congruence index=0.54] and a larger increase in knowledge about ACP (ES=0.22). Intervention patients also experienced response shifts in values and conceptualizations of QOL, becoming less willing to undergo life-sustaining treatments for a new serious medical problem (ES=0.46) or for an incurable progressive disease (ES=0.24), and less willing to tolerate poor health states (ES=0.79). There was no detectable shift in internal standards. We conclude that an intervention of facilitated discussion may improve the congruence of patient wishes and EOL care, and may help define patient values and conceptualizations of QOL at EOL. Consequently, ACP should be an iterative process that is revisited as life experiences and one’s understanding of the benefits and burdens of life-sustaining treatments change over time.
47/ USING ITEM RESPONSE THEORY (IRT) AND STRUCTURAL EQUATION MODELING (SEM) TO INVESTIGATE RESPONSE SHIFTS
Frans J. Oort, Pythia T. Nieuwkerk, Mirjam A. Sprangers, Medical Psychology, Academic Medical Centre, University of Amsterdam, Amsterdam, The Netherlands
When assessing change in quality of life (QOL), we must account for response shifts such as reconceptualization, reprioritization, and recalibration. To detect response shifts we use the IRT&SEM approach, i.e., (1) find appropriate measurement model, using generalized linear IRT, (2) investigate changes in model parameters, using SEM, (3) interpret findings, using a change typology. We illustrate this approach with QOL data. HIV infected patients (216) completed the MOS-HIV Health Survey at the start of antiretroviral therapy and 36 weeks later (T2). The survey measures patient’s health status (PHS) with 11 scales. We used 7 scales: general health (GH), social function (SF), energy/fatigue (EF), quality of life (QL), cognitive function (CF), mental health (MH), and health distress (HD). We applied the IRT&SEM approach with (1) a linear model with 2 common factors, named overall GH (OvGH) and overall MH (OvMH), (2) likelihood-ratio tests to test for invariance of model parameters (e.g., factor loadings and means, variances, and correlations of common and specific factors), and (3) the response shift typology to interpret parameter changes. We found no significant differences between factor loading patterns or values, so neither reconceptualization of OvGH or OvMH, nor reprioritization of the 7 underlying scales occurred. We did find a higher T2 specific factor mean of the HD scale indicating recalibration. At T2 we also found: lower specific factor variances indicating that the patient group became more homogeneous in HD, CF, and SF; a higher correlation between OvGH and OvMH indicating reconceptualization of PHS; and higher common factor means indicating general improvement of OvGH and OvMH. Application of the IRT&SEM approach is useful because it reveals different types of change that would remain hidden otherwise. Underlying assumptions of the IRT&SEM approach, and caveats, are subject of discussion.
48/ RESPONSE SHIFT IN THE ASSESSMENT OF HEALTH RELATED QUALITY OF LIFE (HRQL) POST-STROKE
Sara Ahmed, Nancy E. Mayo, Sharon Wood-Dauphinee, Jim Hanley, Epidemiology, McGill University, Montreal, QC, Canada
The validity of assessing changes in HRQL over time is threatened if people experience a response shift. This study assessed whether the recovery process following stroke altered individuals’ perceptions of past health status and the impact that change in internal standards had on rating of HRQL over time. Individuals were recruited through a randomized trial of acute post-stroke care. Health status was evaluated at baseline and 6 and 24 weeks later using the thermometer scale of the EQ-5D. At 6 and 24 weeks, subjects were asked to retrospectively re-evaluate their health status for the preceding evaluation using the then test technique. Retrospective rating of baseline health status was 10% lower than the prospective rating (p <0.0001) and at 6 weeks, 13% lower (p <0.0001). Measured prospectively, mean health status significantly improved from 67 (/100; SD: 19) at baseline to 76 (SD: 17) at 6weeks. When baseline health status was retrospectively measured, it was re-evaluated at 60 (SD:20) and correspondingly there was an increase in the magnitude of change. Prospectively, between 6 and 24 weeks, there was no measured change in health status. When the retrospective re-evaluation of the 6-week rating was incorporated, there was a significant improvement between 6 and 24 weeks. Memory had an impact on response shift; subjects with "poor" memory had greater response shift and variability. The results suggest that there was a change in internal standards, and that measures of improvement in health status are different based on prospective versus retrospective ratings. Further understanding of the impact of memory on the assessment of response shift using the then test will help to validate the use of this technique.
49/ RESPONSE SHIFT IN PATIENTS' LIVING WITH ADVANCED CANCER
Nessa Coyle, Richard Payne, Neurology, Memorial Sloan-Kettering Cancer Center (MSKCC), New York, NY
Clinical work with patients living with advanced cancer involving palliative care and psychological and social support has led to the observation that at least in some individuals, a response shift occurs in what they consider to be an acceptable QOL. The purpose of this research was to identify, through listening to a series of personal narratives, evidence of this response shift and influencing factors. This is a qualitative exploratory study on the experience of 7 patients living with advanced cancer. Adult English speaking patients with advanced disease, who were followed by the Pain and Palliative Care Service at MSKCC, had expressed attitudes towards life and death to their physician or nurse, and were not known to the principle investigator from her clinical practice were eligible for the study. A series of from 2-6 open-ended audio taped interviews were conducted with each participant either in the hospital, clinic, terminal care facility or their home. Time between each interview varied from 24 hours to a week. Five patients died within one month and two within one week of their last interview. Thematic analysis of their narratives was used. We are developing an indepth sytem to track processes and changes. An illustrative case of response shift was seen in a 60-year old man who was interviewed 6 times over one-month. He spoke of how he came to view care from his wife and sons not as an overwhelming burden that had initially led him to desire a hastened death, but as a positive aspect of their relationship and as a "gift." He also changed the way he saw his progressive illness from one of fighting to one of acceptance. Initial analysis of other participants' narratives illustrates the impact of the environment, relationships and forms of communication on shifting responses to progressive illness. This study highlights the potential for rapid and potentially radical swings in the way terminally ill patients appraise QOL, indicating the need to take response shifts into account in any study with this population.
50/ SYMPTOM EXPERIENCE AND HEALTH RELATED QUALITY OF LIFE AFTER IMPLANTATION OF A CARDIOVERTER DEFIBRILLATOR
Susan F. Marden, Clinical Center Nursing Department, National Institutes of Health, Bethesda, MD
The efficacy of implantable defibrillator (ICD) therapy is well established, yet the impact of recipients' appraisals of side effects associated with disease and treatment is uncertain. This study examined whether symptom experience (SyE) alters Health related Quality of Life (HRQL) in ICD recipients ( n =117). The Important Concerns and Distress Questionnaire (ICDQ) and Short Form-36 (SF-36) were completed by mail survey. SyE is a combination of perceived frequency and associated distress of 29 concerns on the ICDQ. Recipients ( 81% male; 62 years; 3.9 years post ) had experienced at least 1 shock (57%) and cardiac arrest (50%). Most (73%) reported >10 concerns with women reporting more than men (t = 1.79; p < .05). Frequent concerns were also most distressing: tiredness, less sexual activity, trouble sleeping, ICD awareness, shortness of breath, forgetfulness, frustration, weakness, future anxiety, dependency, shock worry. Factorial ANOVA ( SyE-high/low; age <65 or >65; years post <2, 2-5, >5) indicated a main effect for SyE on mental and physical component summary scores of the SF-36 (MCS: F =35.9, p< 05; PCS: F =15.7, p< 05). High SyE levels were associated with a significant, independent reduction in both mental and physical HRQL. Interventions to ameliorate symptom experiences of ICD recipients are warranted and may improve HRQL.
51/ PREDICTING UTILIZATION AND COSTS IN A ONE-YEAR FOLLOW-UP OF HEART FAILURE PATIENTS: THE CONTRIBUTIONS OF GENERIC AND DISEASE SPECIFIC HRQOL ASSESSMENTS
Deborah M. Miller, Mellen Center, Neurology, Cleveland Clinic Foundation, Cleveland, OH; D. Nelson, M. Karafa, Biostatistics and Epidemiology, J. Michaels, Center for Clinical Outcomes, J. B. Young, Cardiology, CCF, Cleveland, OH
The purpose of this study was to determine the factors, including generic and disease-specific HRQoL measures, that contribute to unscheduled health care utilization anywhere and treatment costs at a tertiary care center for cardiac patients who were followed for a maximum of 14 months. Patients who gave informed consent completed two HRQoL measures, the MOS SF-36 and the disease specific Minnesota Living with Heart Failure (LwHF) and reported unscheduled health care utilization (HCU) for the preceding 6 months. They agreed to provide phone follow-up at 6-month intervals for the next year. Objective measures of disease severity, including New York Heart Association classification, type of cardiac disease and medications were retrieved at baseline only. 794 patients provided baseline date. Using a professional survey organization, 181 of these patients were successfully contacted and provided phone follow-up once (either between 6 to 8 months or between 12 to 14 months) and 60 provided follow-up at both assessment phases. When combining all available follow-ups, among individuals who had not had previously reported HCU, either recent LwHF or change in LwHF was significantly higher for those with recent HCU. However, LwHF scores did not differ between those with and without HCU in the subgroup with a history of HCU. Logistic regression was used to compare the relationship of LwHF and the SF-36 with recent HCU. For each of the components of HCU, previous HCU was most strongly associated with recent utilization. Among the survey instruments, LwHF had the strongest association with recent hospitalization, and SF-36 scores had the strongest associations with ED visits and unscheduled medical visits. Average cost of care during one-year study period was $25,449. Of baseline data, SF-36 PCS was most highly correlated with this cost (r=-0.17, p<0.001).
52/ LONG TERM QUALITY OF LIFE AFTER OPEN HEART SURGERY
Djordje Jakovljevic, Zoran R. Potic, Bogoljub Mihajlovic, Marijela M. Potic, Ninoslav D. Radovanovic, Cardiovascular Surgery, Institute of CVD, Sremska Kamenica, Vojvodina, Yugoslavia
To estimate quality of life as an outcome of open heart surgery in long term follow-up (five years). For assessment of quality of life, we used diseases and treatment specific health related quality of life questionnaire and derived integral numeric QOL index QOLi-NS (zero value represents complete inability for any functioning and value 100 means perfect health). The prospective study involved 793 patients who were interviewed preoperatively. Until now, 165 of them completed all of six postoperative controls (6, 12, 24, 36, 48 and 60 months after the operation). Average QOLi-NS before the operation was 44.1. Six months after the QOLi-NS was 83.8, after one year 87.2, after 2 years 85.6, after three years 84.4, after four years 77.9 and 77.8 after five years. Statistically (and clinically) significant difference (p<0.001) in QOL was observed comparing QOL before and after open heart surgery. The highest value of QOLi-NS was one year after the surgery. Quality of life of cardiac surgery patients (average QOLi-NS) increase significantly after the operation and remains very stable till fourth year. At that time it starts to decline, but still is significantly much better than before the treatment.
53/ DETERMINANTS OF THE QUALITY OF LIFE OF PATIENTS WITH CONGENITAL HEART DISEASE
M. Rose, K. Köhler, Psychosomatic, F. Köhler, Cardiology, Charité Humboldt University, Berlin, Germany
What factors determine the health-related quality of life of patients with congenital heart defects? 111 patients (aged 33±12 years) with different degrees of cardiac dysfunction were examined (NYHA 0: 2 I: 56, II: 38, III: 13, IV: 2). All patients underwent a treadmill ergometry in order to determine their level of cardiopulmonary functioning (peak oxygen consumption: VO2max) and answered established questionnaires concerning their quality of life (WHOQOL-Bref), their cardiac complaints (Giessener Complaint Questionnaire GBB), their personality traits (Giessen Test GTS), and their social support (Social Support Questionnaire SOZU-k22). The data was analysed by means of a linear structural equations model. The model proposed was valid, showing good indices of fit (chi2=1.18 p=.55, GFI=.996). It could be shown that the level of cardiopulmonary functioning was most significant for the reporting of specific cardiac complaints (b=-28) and for the physical component of the general, health-related quality of life (b=32), although the former was also influenced by a depressed disposition (b=-.20) and the extent of social support experienced (b=.18). The objective findings, however, had virtually no individual significance for the psychological (b=.09) and social domains (b=-.02) of the HRQL. These HRQL domains are primarily influenced by depressive personality traits (b=-26/-.16) and the social support experienced (b=.51/.51). All three predictors have individual significance for the global assessment of the quality of life (b=.12/.-15/.28). In general, there is a small bond between subjective complaints and objective findings, suggesting that it is necessary to look for factors which influence the expression of specific complaints and other components of the quality of life construct. In the sample studied here, we found that a patient’s depressed disposition and his/her reported experience of social support had added significance for the understanding of the subjective well-being of the patient.
54/ PSYCHOMETRIC PROPERTIES OF THE MACNEW HEART DISEASE QUESTIONNAIRE IN PATIENTS AFTER MYOCARDIAL INFARCTION
Stefan Höfer, Medical Psychology and Psychotherapy, University of Innsbruck, Innsbruck, Austria, Werner Benzer, Working Group on CR & SP, Austrian Society of Cardiology, Feldkirch, Austria; Dieter Brandt, Working Group on CR & SP, Austrian Society of Cardiology, St. Radegund, Austria; Herbert Laimer, Working Group on CR & SP, Austrian Society of Cardiology, Bad Tatzmannsdorf, Austria; Peter Schmidt, Working Group on CR & SP, Austrian Society of Cardiology, Bad Schallerbach, Austria; Arthur Bernardo, Clinic of Cardiac and Psychosomatic Rehabilitation, Gais, Switzerland
Health-related quality of Life (HRQL) represents an important aspect within the assessment of medical treatment. Myocardial infarction (MI) is a major life event and impairs HRQL. After MI important restrictions can be found in physical, emotional and social aspects. For the evaluation of the results of treatment and rehabilitation, appropriate instruments are required. The purpose of this study was the psychometric testing of the translated and linguistic-validated MacNew Heart Disease Questionnaire (MacNew) in MI-patients. 199 subjects (161 male, 38 female) with documented MI received the self-administered MacNew Heart Disease Questionnaire and the SF-36 as the "gold standard" at the beginning of the rehabilitation period. The MacNew identifies disease specific physical, emotional, social and global HRQL. Test-retest reliability, Cronbachs alpha, convergent and discriminate validity coefficients were calculated. Item response rates were >92% for the MacNew. Test-retest reliability [4 weeks] ranged between rtt= 0.72 - 0.87. Cronbachs alpha between 0.89 - 0.93. Convergent validity with the SF-36 scales ranged for the questionnaire from r= 0.36 - 0.75 (p<.001). Discriminate validity established by different grades of severity was high for all scales (ANOVA p<.01). Our results show, that the German MacNew is a reliable and valid instrument to measure HRQL in patients with MI. It enables to discriminate between different grades of severity. Therefore it can also be considered to be sensitive. The high item response rate shows the acceptance of this questionnaire by the patients. The German MacNew can be recommended for evaluative purposes in patients with MI.
55/ THE FACTOR STRUCTURE OF THE SF-36 PHYSICAL AND MENTAL HEALTH SUMMARY MEASURES BY RACE/ETHNICITY AND LANGUAGE IN A US PATIENT POPULATION
Leo S. Morales, Department of Medicine, Steven P. Reise, Department of Psychology, Ron D. Hays, Department of Medicine, University of California, Los Angeles, CA
Leo S. Morales, Department of Medicine, Steven P. Reise, Department of Psychology, Ron D. Hays, Department of Medicine, University of California, Los Angeles, CA
The growing diversity of the US population was made evident by the 2000 census. Until recently, few researchers have given attention to the equivalence of self-report measures among subgroups within the US population. In this study, we examine the factor structure of the SF-36, in four racial/ethnic subgroups: Whites, Blacks, Asians and Latinos. Subjects: A random sample of patients receiving medical care from a medical group association in United States was studied. A total of 7,093 questionnaires were returned for an overall response rate of 59%. Approximately 80% of patients completing surveys self-identified as White (n = 5,508), 10% self-identified as Latino (n = 713), 4% as Asian (n=298), and 3% as Black (n=210). Latinos were further divided among who those responded using a Spanish questionnaire (n=181) and English questionnaire (n=532). SF-36 Health Survey: The SF-36 is a generic HRQOL measure that consists of 36 items, yielding 8 scales (physical functioning, bodily pain, role limitations due to physical health problems, role limitations due to emotional problems, emotional well-being, social functioning, energy/fatigue, and general health perceptions) as well as summary physical and mental health measures. The 8 SF-36 scales were factor analyzed using a full information factor analysis program designed to handle ordinal data. Based on several number of factor criteria, two factors were extracted and rotated using the Promax rotation method. The general health scale score loaded primarily on the physical health factor among Whites, Blacks, and Latinos-English and primarily on the mental health composite among Asians and Latinos-Spanish. The role-emotional scale loaded primarily on the mental health factor among Whites, Asians and Latino-Spanish, and loaded approximately equally on both factors among Blacks and Latino-English. The energy/fatigue scale loaded primarily on mental health among Whites, Asians, Latinos-English and Latinos-Spanish and loaded approximately equally on the two factors among Blacks. Finally, the social functioning scale loaded primarily on physical health among Asians, Blacks, Latino-Spanish and primarily on the mental health factor among Latino-Spanish. The social functioning loaded approximately equally on the two factors among Whites. In general, the hypothesized 2-factor model fit the responses to the SF-36 items very well for all racial/ethnic and language subgroups. However, the pattern of factor loadings varied significantly among the groups. These results have implications for scoring of summary scores from the SF-36. It is possible that distinct scoring protocols are needed for racial/ethnic and language subgroups represented within the United States.
56/ VALIDATION OF THE WHOQOL-BREF IN A NORWEGIAN GENERAL POPULATION
Berit R. Hanestad, Department of Public Health and Primary Health Care, University of Bergen, Bergen, Norway; Astrid K. Wahl, Faculty of Health, Buskerud College, Drammen, Norway; Tone Rustoen, Faculty of Nursing, Oslo University College, Oslo, Norway; Anners Lerdal, Øistein Knudsen, Faculty of Health, Buskerud College, Drammen, Norway
WHOQOL-BREF is a quality of life questionnaire developed for cross-cultural adaptation. The purpose of this study was to validate the questionnaire in a general Norwegian population. 4000 randomly assigned people were sent the questionnaire by mail and 1867 returned it after one reminder. 48% were female and mean age was 45 years(SD 15.9 years). WHOQOL-BREF contains a total of 26 items in relation to the physical health (7items, alpha.83), psychological (6items,alpha.82), social relationships(3items,alpha.60) and environmental domain (8items,alpha.79) and two overall items measuring overall quality of life and general health. Domain score range from 4-20. Higher scores indicate better quality of life. Mean domain scores were 15.8 (SD 2.8), 15.2 (SD 2.4), 15.0 (SD 2.7) and 15.2 (SD 2.4) for the physical health, psychological, social relationships and environmental domain respectively. The results showed that women scored significantly lower within the social relationships domain (p=.005) and higher within the psychological domain (p<.000). Higher age was negatively associated with quality of life within the physical health domain (p<.000) and positively associated within the environmental domain (p<.000). Higher educational level was associated with increased quality of life within the physical health, psychological and environmental domain (p<.000). The questionnaire discriminated in all domains between those who were reporting having a chronic illness compared to those who did not (p<.000) and those living alone compared to those living together with either a spouse or children (p<.000). Correlation analysis between overall quality of life and the different domains ranged from .52 to .65 ( p<.000) for the social relationships and psychological domain respectively and between overall health and the different domains the correlation ranged from .34 to .69 (p<.000) for the social relationships and physical health domain respectively. Principal component factor analysis with varimax rotation resulted in a four factor solution explaining 53% of the variance. The rotated matrix partly confirmed the domain structure. In conclusion the questionnaire seemed to be a valid instrument discriminating different subgroups as expected.
57/ HOW WELL DO HEALTH STATE VALUATION TECHNIQUES CROSS CULTURES? IMPLICATIONS OF CONCEPTIONS OF DEATH AMONGST THE KAMBA OF KENYA
Charles Nzioka, Sociology, University of Nairobi, Nairobi, Kenya; Julia A. Fox-Rushby, Public Health and Policy, London School of Hygiene and Tropical Medicine, London, UK; Patrick I. Muia, Anthropology, University of Nairobi, Nairobi, Kenya
Valuation techniques developed in the USA and Europe are being used around the world either to revalue an existing descriptive system or to weight newly developed questionnaires. Whilst the cross-cultural applicability of HRQL measures is being questioned,