MOS-HIV Health Survey. Users Manual

Transcription

1 MOS-HIV Health Survey Users Manual Albert W. Wu, MD, MPH Johns Hopkins University 1996, 1999 Albert W. Wu All rights reserved MOS-HIV Health Survey Manual

2 Overview Quality of life measures are increasingly being recognized as important when comparing the efficacy of AIDS therapies and assessing the impact of HIV/AIDS on peoples lives. An important question in all HIV/AIDS evaluations is how the virus/disease, medications and other treatment regimens affect patients quality of life. The purpose of this manual is to present the procedures involved in measuring patients quality of life using the Medical Outcomes Study HIV Health Survey (MOS-HIV) and to provide a reference for questions that may arise. The most effective way to use this manual is to first read it over and familiarize yourself with the MOS-HIV. This initial reading will introduce you to background information, general interviewing principles and the details of administration. Background Overview of Quality of Life Measurement What is it? Quality of life is a concept about which everyone has an intuitive understanding, but which is still difficult to define. Quality of life can be affected by many factors including income, housing situation, social interactions, and health. In clinical trials, we are particularly interested in how medical or pharmacologic interventions affect health-related quality of life. There is general agreement that health-related quality of life includes at least a person's physical, social, and cognitive functioning and subjective sense of well-being. Why is quality of life important? It is essential to understand the impact of a disease on many aspects of patients lives. Furthermore, it is important that the effects of drug treatment and other medical regimens translate into benefits that patients can experience -- that the drug or treatment will help them function and feel better. Although traditional outcomes, such as mortality, physiologic changes, and adverse events provide useful clinical and biological information, they may not accurately represent the effect of treatment on the patient's physical, psychological, and social functioning and their subjective sense of well-being. In addition, some of the effects of a drug or treatment may be in areas not accessible to physiologic measurement, and may only be evaluated using patients' assessments. These areas might include effects on energy, pain, or generally how patients feel. Although treatments frequently influence these parameters, a clinical trial is unlikely to demonstrate effects in these areas if patientreported outcomes are not included. 2

3 In AIDS clinical trials, patient-reported health status may be especially important when all the therapies being tested are expected to result in physiologic improvement and survival. In these studies, the most important difference in treatment outcome may be the patients experience. Quality of life scales have the additional advantage of being able to integrate the positive effects of treatment on disease and the negative effects of treatment to give "net" effects in measurable areas. Without such measures, it can be difficult to interpret if a drug that both causes some symptoms and prevents others has a beneficial net effect. For example, it is difficult for a clinician to decide whether a patient should take a drug that is known to cause headache and anorexia, but decreases fatigue. Ultimately, knowledge of drug or treatment effects on patients' quality of life will allow clinicians to tell patients about the anticipated effects of a treatment in terms they can understand. How do we measure it? In general, quality of life is measured by asking a series of questions about specific aspects of functioning and well-being. Asking a series of questions about a particular aspect, such as mental health, allows for a better approximation of the person's emotional state. For example, the answers to five questions about mental health can be added up to give an indication of the person's mental health. Over the past twenty years, significant advances have been made in quality of life measurement. There are a number of well-validated instruments, some of which are comprehensive and intended for use as outcome variables in research studies, and others which are shorter and may be more useful in multicenter clinical trials (Kaplan, Bergner, Brook, Stewart). The MOS-HIV Health Survey traces its genealogy back to several of these well-tested questionnaires, particularly those developed in the late 1970's for the RAND Health Insurance Experiment, and the mid 1980's for the Medical Outcomes Study. Advantages of patient-reported data on quality of life. Patient reported data has a number of unique advantages over other sources of data. Most importantly, this form of data gives researchers access to information that is not available from any other source, and which is of great importance to the patient. Without much thinking, it is obvious that the most relevant and valid information about ability to function and quality of life must come from the person her/himself. This position is supported by the fact that both clinicians' and family members' estimates of functioning are less reliable, and often do not agree with those of the patient (Sprangers, Wu and Jacobson). (Of course, in those cases where the patient's viewpoint is not known and cannot be obtained, a surrogate's estimate of the patient's functioning and well-being may be used if there is reason to believe that the surrogate validly represents the patient's views.) Disadvantages of patient-reported data on quality of life 3

4 To a greater extent than with data from other sources, missing Quality-of-Life data is precisely the data that you are most interested in having. For example, a patient who misses a visit for a reason related to health is also likely to have worse-than-average quality of life. Therefore, it is important to incorporate strategies to minimize missing data, to document reasons for missing data, and to obtain answers from reliable surrogates when possible. Patient reports of their quality of life are by their nature subjective. Patient- reported data can differ from data obtained on the patient from other sources. For example, in general patients tend to be more optimistic about their abilities than family members or physicians. Or they may judge their health to be better than others would. Patient reported data can also be idiosyncratic. There are sources of variability in patients' responses that are not related to the treatment they are receiving. In addition, some aspects of patient's quality of life can be influenced by things other than their illness or treatment. For example, if a person gets evicted from their apartment, their mood may be depressed. This is a source of legitimate variability in a person's quality of life, and will be reflected in the answers they give to the questionnaire. However, patient-reported responses are legitimate perceptions of what the patient thinks or feels. Fortunately, we have the benefit of randomization in clinical trials, and there is no reason to believe that such non-health related events will be unevenly distributed between treatment groups. The MOS-HIV Health Survey The Medical Outcomes Study HIV Health Survey (MOS-HIV) is a brief, comprehensive measure of health-related quality of life used extensively in Human Immunodeficiency Virus/Acquired Immune Deficiency Syndrome (HIV/AIDS). Questions in the MOS-HIV were drawn from a large pool of existing questions that had been extensively tested for use in the Medical Outcomes Study, a large multisite study of the effects of different ways of delivering medical care. Developed in 1987, the MOS-HIV was one of the first disease-targeted measures available for this populations. The survey is widely used in clinical trials and other research and evaluation studies. The 35-item MOS-HIV assesses aspects of functioning and well-being including physical function, social and role function (work), cognitive function, pain, mental health, energy, distress about health, quality of life and overall health. (Wu, Rubin, Mathews, et al., 1991) 4

5 Development of the MOS-HIV Health Survey The development of the MOS-HIV was prompted by the need for a brief instrument to assess aspects of functional status and well-being in persons with HIV disease for use in multi-center AIDS clinical trials. Based on scales developed from the Medical Outcomes Study, the dimensions represented in the SF-20 (Stewart, Hays and Ware, 1988) served as the basis of the instrument. Discussion and input from clinic patients, clinical trial participants and providers of care to HIV/AIDS patients led to the addition of concepts important in the disease. Measures of energy/fatigue, cognitive functioning, health distress and quality of life were added from the MOS item pool (Stewart and Ware, 1992) to create a 30-item version of the survey. (Wu, Rubin, Mathews, et al., 1991) In the initial MOS-HIV instrument, also referred to as the MOS-30, a single overall health item (Manning, Newhouse and Ware, 1991) was used as a measure of general health perception. Due to the importance of this concept, four additional items were added to increase reliability. This resulted in the MOS-34. Because the pattern of scores from the SF-20 (Stewart, Hays and Ware, 1988), general health perception scale more closely approximates a normal distribution in HIVinfected populations, the four items were chosen from the SF-20 rather than the SF-36 (Ware and Sherbourne, 1992) which had a skewed distribution on general health perception. Since pain is a prevalent symptom among patients with HIV/AIDS (O Neill and Sherrand, 1993) and a common adverse effect of some treatments, it is important to assess this concept with greater precision than would be afforded by a single item. Therefore, a second pain item from the SF-36 was added in 1991 for use in a large international clinical trial. (Feinberg, Cooper and Hurwitz, 1996) At the same time, response choices were altered for physical functioning to reflect extent (limited a lot, limited a little, not limited) rather than duration of limitation (for more than 3 months, for 3 months or less, not limited). The response categories also changed for role functioning to reflect presence or absence of limitations. Summary of Concepts and Measures The MOS-HIV Health Survey consists of 35 questions which assess 10 dimensions of healthrelated quality of life including general health perceptions, physical functioning, role functioning, pain, social functioning, mental health, energy, health distress, cognitive functioning, quality of life. (see Table 1) In addition, one item assesses health transition. The subscales of the MOS-HIV are scored as summated rating scales on a scale where higher scores indicate better health. For multi-item scales (two or more items), mean substitution is generally used for missing items if no more than 50% of the items are missing. In addition to these subscales, a Physical Health Summary score and a Mental Health Summary score can be generated. The instrument takes approximately 5 minutes to complete and can be self-administered using paper and pencil or a touch screen personal computer or administered as a face-to-face or telephone interview. A brief description of each dimension is presented below. General Health Perceptions - This dimension utilizes a five-item scale adapted from the SF- 5

6 20 which used a single-item rating of health and four items from the Current Health scale from the Health Perceptions Questionnaire. (Davies and Ware, 1981; Ware, 1976) This dimension retained the five items measuring current health from the SF-20. Individual items in the scale ask patients to report on their general health, resistance to illnesses and health outlook. Davies and Ware (1981) and Stewart and Ware (1992) have reported substantial empirical evidence of validity for this scale. Physical Functioning - This dimension consists of six items that assess a range of severe and minor physical limitations. These items represent different levels and kinds of limitations including lifting heavy objects or participating in strenuous sports, walking uphill or climbing a few flights of stairs, bending, lifting or stooping, and walking a short distance. Limitations in self-care activities are measured with a single item assessing the ability to eat, dress, bath or use the toilet by oneself. 6

7 Table 1. Summary of MOS-HIV Survey Concepts Meaning of Scores Concepts No. of Items Low High General Health Perceptions 5 Views personal health as poor Views personal health as excellent Physical Functioning 6 Very limited in performing physical activities due to poor health including eating, dressing, bathing or using the toilet Performs all types of physical activities due to poor health including vigorous or strenuous activities without limitations Role Functioning 2 As a result of physical health, experiences problems with work or daily activities No problems with work or other daily activities as a result of health Pain 2 Very severe and limiting pain No pain or limitations due to pain Social Functioning 1 Social activities limited due to health No limitations on social activities as a result of health Mental Health 5 Feels nervous and depressed all of the time Feels calm, peaceful and happy all of the time Energy 4 Feels tired and worn out all of the time Feels energetic and full of pep all the time Health Distress 4 Feels despair, discouraged and afraid due to health all of the time Does not feels despair, discouraged and afraid due to health Cognitive Functioning 4 Has difficulty concentrating, reasoning and remembering all of the time Has no problem concentrating, reasoning and remembering Quality of Life 1 Life has been very bad; could hardly be worse Life has been very good; could hardly be better Health Transition 1 Physical health and emotional condition much worse than 4 weeks ago Physical health and emotional condition much better than 4 weeks ago Note: Adapted from The MOS 36-Item Short-Form Health Survey (SF-36). I. Conceptual framework and item selection by J.E. Ware and C.D. Sherbourne, 1992, Medical Care, 30: The response categories permit estimation of the severity of each limitation. Early quality of life surveys measured the duration of any reported limitation. However, most physical limitations are chronic in nature and thus, measures of duration are of little value for data analysis. (Stewart et al., 1981) Analytical precision is increased with the ability to distinguish between patients who are able to perform physical activities with and without some degree of difficulty. (Stewart and Kamberg, 1992) Like the SF-36, the MOS-HIV Health Survey utilizes a three-level response continuum that measures both the presence and degree of physical limitations. (Ware, 1993) 7

8 Role Functioning - Two items are used to assess the impact of patients health on their ability to perform on the job, around the house or in school. Patients are asked if their health keeps them from working at a job, doing work around the house or going to school. The second item in this scale asks if patients are unable to do certain kinds of work, housework or schoolwork because of their health. Pain - Using questions similar to those found in the SF-20 and SF-36, the MOS-HIV Health Survey assesses both the intensity of bodily pain and the degree of interference with normal activities due to pain. (Ware,1993) Social Functioning - This single item subscale asks patients to assess the extent to which their health in the past 4 weeks has limited their social activities. Precision is increased in the MOS- HIV Health Survey by specifically assessing the impact of patients health on social activities thus eliminating the influence of non-health factors on social activity. (Stewart Hayes and Ware, 1988) Mental Health - The MOS-HIV Health Survey utilizes the same five-item Mental Health scale as found in both the SF-20 and SF-36. One or more items from each of the four major mental health dimensions (anxiety, depression, loss of behavioral/emotional control and psychological wellbeing) are included in the scale. (Veit and Ware, 1983) Items in this scale present a balance between favorably and unfavorably worded items thus controlling for response set effects. Energy/Fatigue - This four-item scale is included in the MOS-HIV Health Survey to measure differences in vitality. As in the Mental Health scale, items in this scale control for response set effects. Health Distress - This dimension assesses the degree to which patients are discouraged and afraid due to their health problems. The four-item scale also asks patients about the amount of time in the past 4 weeks that they have felt despair and weighed down by health problems. Cognitive Functioning - Consisting of four-items, this dimension measures the degree of difficulty patients have experienced in the past four weeks with respect to their cognitive abilities. Patients are asked to assess how much of the time they have had difficulty reasoning or solving problems, been forgetful, had difficulty in remaining attentive and concentrating on activities. Quality of Life - In this single item dimension, patients are asked to assess the quality of their life during the past 4 weeks. The response categories range from very well, could hardly be better to very bad, could hardly be worse. Health Transition - This question asks patients about the amount of change in their physical and emotional health over a four week period. Like a similar item in the SF-36, the health transition question has been found to provide useful information about actual changes in health status during the period prior to the administration of the MOS-HIV Health Survey. (Ware, 1993) Physical Health and Mental Health Summary Scores 8

9 Using the ten subscales of the MOS-HIV, both physical health and mental health summary scores have been developed. (Revicki, Sorensen and Wu, 1998) Physical health and mental health factors were derived from exploratory and confirmatory factor analyses of MOS-HIV data from over 2,500 patients with HIV disease. The use of summary index scores rather than multiple scale scores simplify data analysis and the interpretation of findings from clinical trials. These summary scores also aid in comparisons across studies. (Revicki, Sorensen and Wu, 1998) For the Physical Health Summary (PHS) score, the physical function, pain, and role function scale scores contributed most strongly. For the Mental Health Summary (MHS) score, the mental health, health distress, quality of life, and cognitive function scales contributed most strongly. The vitality, general health and social function scales contributed to both factors. The summary scores are transformed to t-scores with a mean of 50 and a standard deviation of 10. In an analysis of these measures, Revicki et al. (1998) found differences in the mean PHS and MHS scores among patient groups defined by HIV disease stage and severity, Karnofsky performance status scores and global ratings of health status. Patients reporting worsening health status had significantly lower mean PHS and MHS scores than patients reporting stable or improving health status. These measures are reproducible results across different samples of HIV/AID patients. (Revicki, Sorensen and Wu, 1998) The Scoring Algorithms for the PHS and MHS scores are presented in this manual. Administration of the MOS-HIV Health Survey Methodological issues The following section provides guidelines for administering the MOS-HIV Health Survey. The instrument can be self-administered or a trained interviewer can conduct a telephone or face-to face interview with patients. The instrument takes approximately five minutes to complete. The MOS-HIV Health Survey is most frequently completed by patients in a clinical setting. However, the instrument can also be completed as a mailed survey or a telephone or in-person interview in other settings, such as the home. The MOS-HIV can also be included as one section of a longer interview or questionnaire. More recently, computer assisted administration has been introduced using touch screen technology. A study coordinator or someone in a similar role who facilitates the administration of the MOS-HIV Health Survey plays a crucial role in any data collection effort. The quality of the data will be determined by the skill and effectiveness of the individual assuming this role.. Unlike lab data, the quality of questionnaire data depends in part on setting an appropriate context -- setting the stage -- for the study participant. If a study participant appreciates the importance of the data being collected, considers the questions carefully, and answers appropriately, the responses will be a more accurate reflection of the participant's self-perceived health and quality of life. Study participants will also be more likely to complete the questionnaire if they feel that it is important. When the following procedures are implemented during administration of the survey, the 9

10 quality of data obtained from study participants will be improved. Introduce and explain the survey to study participants When handing out the questionnaire, explain how it is to be completed Collect and review the returned questionnaire for completeness Complete face-to-face interviews if necessary It is important to be familiar with the content and format of the questionnaire before giving the questionnaire to study participants. A thorough knowledge of the instrument will make it easier to answer study participants' questions about the questionnaire, and to edit completed questionnaires for any errors made by participants in filling out the questionnaire. Reading ability and the ability to think abstractly vary among respondents. As such, subjects may vary in the extent to which responses to questionnaires are consistent and valid reflections of their subjectively experienced health and well being. (Ref). (WHAT REF??) The MOS-HIV is written to correspond to an 8 th grade reading level. Thus, subjects who are at least 12 years of age with the corresponding reading level should be able to complete the questionnaire. (Ware, 1993; Stewart and Ware, 1992) Patients who are functionally illiterate or perform below this reading level will require interview administration. Patients who are visually impaired may require a low-vision version of the questionnaire, or interview administration. For example, in the SOCA CMV retinitis retreatment trial (The SOCA Research Group, 1996), all health related quality of life measures were administered as an interview rather than using self-administration because many patients had decreased visual acuity and other impairments and vision might be related to treatment assignment. Furthermore, if there is a high prevalence of severe cognitive impairment, it may be advisable to administer a preliminary mental status examination before asking patients to complete the survey. Patients who are incapable of abstract thinking, or who have other cognitive impairments, may not be able to provide useful response to the questionnaire. Timing of data collection When collecting data in a clinical setting, the MOS-HIV Health Survey should be administered before the patient is seen by a health care provider so that the patient-provider interaction will not influence the patient s survey responses. Further, the survey should be completed before the patient is asked any other questions about their health or illnesses. (Ware, 1993) 10

11 Guidelines for Administering the MOS-HIV Health Survey Administration Protocol The basic steps for self-administration are 1) giving the questionnaire to the study participant, 2) reading the instructions to the participant, 3) answering any questions the participant may have, and 4) leaving the questionnaire with the participant for completion. The best setting for patients to fill out the questionnaire is a quiet secluded area with a minimum of distractions, such as an exam room or other office. It is preferable that patients be separated from family or friends who accompany them to the study visit, so that only the patient's opinions are measured. It is important to explain that the patient s answers are of interest, and that these answers should not be influenced by others. When handing out the questionnaire, the study participant should be told the purpose of conducting the study and given an estimate of the time required to complete the survey. The format of the questions and how to complete them should be reviewed with the patient. All study participants are not equally experienced in filling out questionnaires, or answering multiple choice questions. It should be explained that : all questions are about the patient s level of functioning and well-being over that last 4 weeks. the questions should be answered by placing a check mark in the box next to the response that most closely corresponds to the patient s answer. every question should be answered. some questions may appear similar to others, but each one is different. if the patient is unsure about how to answer a question, he/she should give the best answer possible or write a comment in the left margin. only one answer should be checked, giving the best answer to each question (check only one box) Completed questionnaires should be checked to assure that every question has been answered and that only one box has be checked. 11

12 There are a few situations in which a patient will not be able to fill in the questionnaire on his or her own. Common reasons included lack of reading skills, impaired vision, non-english speaking, extreme malaise, and delirium or dementia. In some cases, a face-to-face interview may be required. Administering a Face-to-Face Interview It is a good idea to practice administering the questionnaire as an interview to other coworkers, family members, or friends. In general, it is important to be prepared when conducting a face-to-face interview. If you are confident and know what you are doing, you will feel more comfortable and this will be perceived by the patient. Practice reading the questionnaire in advance. If you sound like you are reading it, you will get less natural responses. Be willing to answer any questions. If there is a question you cannot answer, explain that you don't know the answer but will find out and call the patient with the answer. Occasionally longer questions may come up. Save these for after the interview. Establishing rapport with the patient is essential for a successful interview. Strive to achieve a good ("friendly but neutral") relationship with the patient. It is important to be sufficiently neutral to avoid biasing the subject's responses. Make sure the study participant is as comfortable as possible before starting. A successful interviewer needs to develop and maintain a comfortable reading pace. Read instructions slowly enough for the subject to understand them (new interviewers have a tendency to read instructions very quickly). Read items slowly enough for the subject to consider each statement and respond. It is also important to try to maintain a reading style that is clear and not monotonous. Try to hold the subject's attention and interest. Clearly emphasize the important words and concepts. Testing of Scaling Assumptions Reliability Data from numerous studies support the internal consistency reliability of the multi-item scales in the MOS-HIV (Table 2). In most cases, Cronbach s alpha coefficients exceed 0.70, suggesting adequate reliability for group comparisons. (Nunnally, 1978) Although in preliminary analysis Cronbach s alpha for the role functioning scale was 0.50 (Wu, Rubin, Mathews, et al., 1991), in subsequent studies coefficients consistently exceeded (Burgess, Dayer, Catalan, et al., 1993; Wu, Lichter, Richardson, et al., 1992; Scott-Lennox, McLaughlin Miley, Mauskopf, 1996; Wu, Jacobson, Grant, et al., 1997; Revicki, Wu, and Brown, 1995; Revicki and Swartz, 1997; Singer, Thorne, Raboud and Shafran, et al., ; Scott-Lennox, Boyer, Ware, and Wu, ; Safrin, Finkelstein, Feinberg et al., 1996; Wu, Gray, Brookmeyer, Safrin, 1996; The SOCA Research Group, 1996; Wu, Jacobson, Berzon, et al., 1997; Zander, Palitzsch, Kirchberger, et al., 1994) 12

13 In initial evaluations, the internal consistency reliability of the physical and mental health summary scores have also been satisfactory. Cronbach s alpha coefficients for the PHS score range from 0.90 to 0.92, and coefficients for the MHS scores range from 0.91 to 0.94 across different samples. (Revicki, Sorensen and Wu, 1998) 13

14 Table 2 Internal Consistency (Cronbach s Alpha) of MOS-HIV Subscales 14

15 Population Asx and Early ARC All Stages AIDS/MAC ACTG 157 CD NUCA 3002 CD NUCA 3001 AIDS CD4<100 ACTG 204 All Stages CD Reference Wu et al., 1991 Burgess et al., 1993 Wu et al, Scott-Lennox et al., 1996 Scott -Lennox, et al., 1996 Wu et al., 1997 Revicki et al., 1995 Revicki et al., 1996 Subscale (N) General health perceptions s s Pain s s s s s Physical functioning Role functioning NA Mental health Energy/fatigue Health distress NA Cognitive functioning Asx = asymptomatic; ARC = AIDS Related Complex; MAC = mycobacterium avium complex infection; s = single item; NA = not available 15

16 Construct validity Multitrait analyses support the convergent and discriminant construct validity of the scales, and suggest that they measure distinct aspects of health across different stages of illness. (Wu, Rubin, Mathews, et al., 1991; Revicki, Wu, and Brown, 1995; Zander, Palitzsch, Kirchberger, et al., 1994; Wu, Jacobson, Clark, et al., 1997a) There is also a large body of data on the relationship of scale scores to other concurrent indicators of health. As expected, MOS-HIV scores are moderately and significantly correlated with scores on established health status instruments, (Wu, Rubin, Mathews, et al., 1991; Burgess, Dayer, Catalan, et al., 1993; Revicki, Wu, and Brown, 1995; Revicki and Swartz, 1997; Hadorn and Hays, 1991; Zander, Jager, Palitzsch, et al., 1993; Doob and MacFadden, 1993; Copfer, Ampel, Hughes, et al., 1996; Campbell, Converse and Rodgers, 1976; Ganz, Schag, Kahn and Petersen, 1994; Schag, Ganz, Kahn and Petersen, 1992; O Leary, Ganz, Wu, et al., ; Fanning, Emmott, Sherett, et al., 1993; Nabulsi, Revicki, Conway, et al., 1996; Wu, Jacobson, Clark, et al., 1997b), symptom indices (Wu, Rubin, Mathews, et al., 1991; Revicki, Wu, and Brown, 1995; Zander, Jager, Palitzsch, et al., 1993), performance measures, (Doob, Johnson, StCyr and MacFadden, 1996; Givertz amd Revicki, 1995) and clinical and examination findings. (Revicki, Wu, and Brown, 1995; Wu, Coleson, Holbrook and Jabs, 1996) (Table 3) People with asymptomatic HIV disease have consistently demonstrated higher scores than patients with later stages of disease. (Wu, Rubin, Mathews, et al., 1991; Burgess, Dayer, Catalan, et al., 1993; Revicki, Wu, and Brown, 1995; Copfer, Ampel, Hughes, et al., 1996; Ganz, Schag, Kahn and Petersen, 1994) (Table 4). Patients with AIDS and those with symptomatic HIV often have comparable scores, although in general, AIDS patients have lower physical health related scores. However, symptomatic HIV-infected patients often have lower mental health related scores. Results from some of these studies could provide a cross-walk from one instrument to another. For example, since the MOS-HIV subscale scores explained 80% of the variance in QWB score, it may be possible to impute QWB scores in studies which employed only the MOS-HIV. (Copfer, Ampel, Hughes, et al., 1996) On the other hand, this would be difficult to do for Standard Gamble utilities, since these scores were weakly related to the MOS-HIV and other variables. (Revicki, Wu, and Brown, 1995) There is also evidence supporting the validity of the physical and mental health summary scores. As expected, physical and mental health summary scores vary in patient groups defined by HIV disease stage, HIV disease severity, Karnofsky Performance Status scores, and global ratings of health status. (Revicki and Swartz, 1997) Physical health summary scores were significantly associated with higher CD4 cell counts. Mean mental health and physical health summary scores for patients reporting a deterioration in health Table 3 Evidence for Construct Validity for the MOS-HIV Concurrent Measures References 16

17 Karnofsky Performance Status scores Wu, Rubin, Mathews et al., 1991 Zander, Palitzsch, Kirchberger et al., 1994 Sickness Impact Profile Revicki, Wu,Brown, 1995 Quality of Well-being Scale (QWB) Coper, Ampel, Hughes, et al., 1996 Campbell s Quality of American Life Campbell, Converse, Rodgers, 1976 HOPES Ganz, Schag, Kahn, Petersen, 1994 Schag, Ganz, Kahn, Petersen, 1992 O Leary, Ganz, Wu, et al, submitted Standard Gamble Utility Revicki, Wu, Brown, 1995 Fanning QOL Scale Fanning, Emmott, Sherett et al., 1993 EuroQol Nabulsi, Revick, Conway et al., 1996 Wu, Jacobson, Clark et al., 1997a Depression scale scores Burgess, Dayer, Catalan et al., 1993 Revicki and Swartz, 1997 Zander, Jager, Palitzsch et al., 1992 Doob, MacFadden, Frequency and severity of physical and mental symptoms Wu, Rubin, Mathews et al., 1991 Revicki, Wu, Brown, 1995 Zander, Jager, Palitzsch et al., minute walk test performance Doob, Johnson, St Cyr, MacFadden, 1996 Cognitive function tests Givertz and Revicki, 1995 Visual acuity Wu, Coleson, Holbrook et al., 1996 Severity of illness (Rabeneck Severity Score) Revicki, Wu, Brown, 1995 status (on the health transition item) were significantly lower than the summary scores of patients reporting stable or improving health status. Mental health summary scores were lower in patients with clinical reports of psychiatric disorders and the mental health summary scores were correlated with scores on the Center for Epidemiologic Study Depression scale. Two studies have demonstrated that the two summary scores predict attrition from clinical trials (Wu, Jacobson, Grant and Scott-Lennox, 1997; Revicki and Swartz, 1997) and are correlated with clinical endpoints and mortality. (Wu, Jacobson, Grant and Scott-Lennox, 1997; Revicki and Swartz, 1997) Responsiveness 17

18 Increasing evidence supports the responsiveness of the scales to clinically important changes. For patients undergoing treatment for Mycobacterium avium complex (MAC) infection, improvements in MOS-HIV scores were related to decreases in MAC bacteremia. (Singer, Thorne, Khorasheh, et al., ; Wu, Brookmeyer, Gray, et al., 1994) For patients with mild-to-moderately severe Pneumocystis carinii pneumonia, improvements in energy subscale scores were related to decreases in A-a gradient (ACTG 108). (Safrin, Finkelstein, Feinberg, 1996; Wu, Gray, Brookmeyer, Safrin, 1996) For patients with advanced disease (CD4 < 100), MOS-HIV subscale scores decreased for patients who experienced adverse events, and for those who developed opportunistic infections during the trial. In a study of saquinavir, Revicki also found a relationship between increased symptoms and lower scores. (Revicki and Swartz, 1997) Cohen and others have also demonstrated that the MOS-HIV subscale scores are related to AIDS-related clinical endpoints. (Cohen, Revicki, Nabulsi, et al., 1997) For the physical health and mental health summary scores, preliminary findings from three completed clinical trials support the responsiveness of the scales. (Revicki, Sorensen and Wu, 1998; Wu, Jacobson, Clark, et al., 1997; Cohen, Revicki, Nabulsi, et al., 1997) Patients with an AIDS-defining event had significantly lower PHS and MHS scores compared to patients with no clinical event. The mean score on the PHS was 48.5 for patients without an AIDSdefining clinical event versus 41.8 for those with a clinical event. While the mean score on the MHS for patients without a clinical event was 49.4, patients with a clinical event scored (Revicki and Swartz, 1997) Scoring of the MOS-HIV Health Survey Standardization When utilizing the MOS-HIV Health Survey, it is important to adhere to the standards of content and scoring outlined in this manual. Changes in either the content of the survey or in the scoring instructions may jeopardize the reliability and validity of scores. Furthermore, changes would prevent comparisons of results across studies. (Ware, 1993) 18

19 General information All items and scales in the MOS-HIV Health Survey are scored so that a higher score indicates better health status. For example, the four functioning scales, physical, role, social and cognitive, are scored so that a higher score is indicative of better functioning. The items and scales of the MOS-HIV Health Survey are scored in three steps: 1. item recoding; 11 of the 35 items in the survey require recoding. If the scale scores are being computed using a SAS program, item recoding can be accomplished in the DATA step by subtracting the item score from the number of items in the scale plus 1, (e.g., ghp = (6 - ghp) + 1; 2. item scores in each scale are summed to compute raw scale scores; and 3. raw scale scores are transformed to a scale (transformed scale scores) to facilitate comparisons with other MOS-HIV Health Survey data. Data Entry Questionnaire responses should be keypunched using the numbers coded on the questionnaire. Item recoding and scale scoring can then most easily be completed using standard data analysis software such as SAS or SPSS. Prior to keypunching, completed surveys should be carefully edited for clarity and accuracy. Solutions for handling common coding problems include: If two adjacent responses are selected by the respondent, randomly select one to be entered. If two non-adjacent responses are selected, code the item as a missing value. If more than one response is selected for a single item, code the item as a missing value. Item Recoding Item recoding is conducted after questionnaire editing and data entry has been completed. This process is completed to derive the item values used to calculate scale scores. The process will require one or more of the following steps: 1) recode values for 11 items, 2) change out-ofrange values to missing, and 3) substitute person-specific estimates for missing items. Recode Values for 11 Items: Eleven items are reverse scored. These items are worded so that a higher precoded item value indicates a poorer health state. To ensure that a higher item value indicates better health on all MOS-HIV items and scale, these 11 items require recoding as detailed in Tables Out-of-Range Values: All 35 individual items should be checked for out-of-range values prior to recoding items to their final item values. Often the result of data entry errors, out-of-range 19

20 values are those that are lower or higher than the item precoded range. If possible, out-of-range values should be changed to the corrected by verifying the response on the original questionnaire. If this is not possible, all out-of-range values should recoded as missing. Missing values: Respondents may fail to complete one or more questionnaire items in a scale. Multi-item scales permit the estimation of a scale score even though some items are missing. A scale score can be calculated if the respondent answered at least half of the individual items in a multi-item scale for those scales consisting of 4 or more items. When an item is missing, substitute the respondent s average score across the completed items in the scale. For example, if a respondent leaves one item in the 5-item mental health scale blank, substitute the average score of the four completed items for the missing value. Computation of Raw Scale Scores The following tables provide scoring instructions for each of the 10 MOS-HIV Health survey scales and the reported health transition item. Each table contains the scale item verbatim from the questionnaire, response choices and both the precoded and final values for scoring each item. 20

21 Table 4. Scoring Information: General Health Perceptions Five items: 1, 11 a-d VERBATIM ITEMS 1. In general, would you say your health is: 11a. I am somewhat ill. 11b. I am as health as anybody I know. 11c. My health is excellent. 11d. I have been feeling bad lately. Item 1, 11b and 11c: Note that values for this items need to be recoded. (If you are programming in SAS, you can accomplish this in your DATA step by subtracting the score from the number of items in the scale plus 1, (e.g., egfp = 6 egfp). SAS program code for scoring the MOS-HIV is included as Appendix D.) Questionnaire Response choice Item Value Final Value Excellent 1 5 Very good 2 4 Good 3 3 Fair 4 2 Poor 5 1 Items 11a and 11d Questionnaire Response choice Item Value Final Value Definitely true 1 1 Mostly true 2 2 Not sure 3 3 Mostly false 4 4 Definitely false 5 5 Scale Scoring: The score from item 1 is summed with the scores for items 11 a-d to form a 5-item General Health Perception scale. The range of scores for this scale before it is standardized is then

22 Table 5. Scoring Information: Physical Functioning Six items: 4a-f VERBATIM ITEMS The following questions are about activities you might do during a typical day. Does your health now limit you in these activities? 4a. 4a. The kinds or amounts of vigorous activities you can do, like lifting heavy objects, running or participating in strenuous sports 4b. The kinds or amounts of moderate activities you can do, like moving a table, carrying groceries or bowling. 4c. Walking uphill or climbing (a few flights of stairs). 4d. Bending, lifting or stooping. 4e. Walking one block. 4f. Eating, dressing, bathing or using the toilet. ITEM SCORING Items 4a-f Response choice Questionnaire Coded Value Final Value Yes, limited a lot 1 1 Yes, limited a little 2 2 No 3 3 Items in the Physical Function scale are simply summed. The range of possible scores for the Physical Function scale before it is standardized is

23 Table 6. Scoring Information: Role Functioning Two items: 5,6 VERBATIM ITEMS 5. Does your health keep you from working at a job, doing work around the house or going to school? 6. Have you been unable to do certain kinds or amounts of work, housework or schoolwork because of your health? ITEM SCORING Items 5, 6 Questionnaire Response choice Coded Value Final Value Yes 1 1 No 2 2 Scores are now summed for the Role Function scale. The range is of scores for Role Function before it is standardized is

24 Table 7. Scoring Information: Social Functioning One item: 7 VERBATIM ITEM 7. How much of the time, during the past 4 weeks, has your health limited your social activities (like visiting with friends or close relatives)? ITEM SCORING Item 7 Questionnaire Response choice Coded Value Final Value All of the time 1 1 Most of the time 2 2 A good bit of the time 3 3 Some of the time 4 4 A little of the time 5 5 None of the time 6 6 The Social Function item is scored as coded before it is standardized with a range from

25 Table 8. Four items: Scoring Information: Cognitive Functioning 10 a-d VERBATIM ITEMS How much of the time during the past 4 weeks: 10a. Did you have difficulty reasoning and solving problems, for example making plans, making decisions, learning new things? 10b. Did you forget, for example, things that happened recently, where you put things, appointments? 10c. Did you have trouble keeping your attention on any activity for long? 10d. Did you have difficulty doing activities involving concentration and thinking? ITEM SCORING Items 10 a-d Questionnaire Response choice Coded Value Final Value All of the time 1 1 Most of the time 2 2 A good bit of the time 3 3 Some of the time 4 4 A little of the time 5 5 None of the time 6 6 Items in the Cognitive Function scale are scored 1-6. None of the items need to be recoded. When the values of the 4 scale items are summed, the range for the Cognitive Function scale before it is standardized is

26 Table 9. Scoring Information: Pain Two items: 2,3 VERBATIM ITEM 2. How much bodily pain have you generally had during the past 4 weeks? 3. During the past 4 weeks, how much did pain interfere with your normal work (including both work outside the home and housework)? ITEM SCORING Item 2 - This item must be recoded. ITEM SCORING Questionnaire Response choice Coded Value Final Value None 1 6 Very mild 2 5 Mild 3 4 Moderate 4 3 Severe 5 2 Very Severe 6 1 Item 3 - This item must be recoded. Questionnaire Response choice Coded Value Final Value Not at all 1 5 A little bit 2 4 Moderately 3 3 Quite a bit 4 2 Extremely 5 1 Note that these items are recoded so that a higher score corresponds to less pain. The range of possible scores for this Pain item before it is standardized is

27 Table 10. Five items: Scoring Information: Mental Health 8 a-e VERBATIM ITEMS How much of the time, during the past 4 weeks: 8a. Have you been a very nervous person? 8b. Have you felt calm and peaceful? 8c. Have you felt downhearted and blue? 8d. Have you been a happy person? 8e. Have you felt so down in the dumps that nothing could cheer you up? ITEM SCORING Items 8a, 8c, 8e Questionnaire Response choice Coded Value Final Value All of the time 1 1 Most of the time 2 2 A good bit of the time 3 3 Some of the time 4 4 A little of the time 5 5 None of the time 6 6 Items 8b, 8d - These items need to be recoded. Questionnaire Response choice Coded Value Final Value All of the time 1 6 Most of the time 2 5 A good bit of the 3 4 Some of the time 4 3 A little of the time 5 2 None of the time 6 1 Items in the Mental Health scale are scored 1-6. The range is 5-30 for the Mental Health scale before it is standardized. 27

28 Table 11. Scoring Information: Energy/Fatigue Four items: 9 a-d VERBATIM ITEMS How often during the last 4 weeks: 9a. Did you feel full of pep? 9b. Did you feel worn out? 9c. Did you feel tired? 9d. Did you have enough energy to do the things you want to do? ITEM SCORING Items 9a and 9d - These items need to be recoded. Questionnaire Response choice Coded Value Final Value All of the time 1 6 Most of the time 2 5 A good bit of the time 3 4 Some of the time 4 3 A little of the time 5 2 None of the time 6 1 Items 9b and 9c Questionnaire Response choice Coded Value Final Value All of the time 1 1 Most of the time 2 2 A good bit of the time 3 3 Some of the time 4 4 A little of the time 5 5 None of the time 6 6 Items in the Energy/Fatigue scale are scored 1-6. The range for the Energy/ Fatigue scale is 4-24 before standardization. 28

29 Table 12. Scoring Information: Health Distress Four items: 9 e-h VERBATIM ITEMS How often during the last 4 weeks: 9e. Did you feel weighed down by your health problems? 9f. Were you discouraged by your health problems? 9g. Did you feel despair over your health problems? 9h. Were you afraid because of your health? ITEM SCORING Items 9e-9h Questionnaire Response choice Coded Value Final Value All of the time 1 1 Most of the time 2 2 A good bit of the time 3 3 Some of the time 4 4 A little of the time 5 5 None of the time 6 6 Items in the Health Distress scale are scored 1-6. None the items in this scale are recoded. The range for the Health Distress scale is 4-24 before standardization. 29

30 Table 13. Scoring Information: Quality of Life One item: 12 VERBATIM ITEM 12. How has the quality of your life been during the past 4 weeks? i.e., How have things been going for you? ITEM SCORING Item 12 - This items needs to be recoded. Questionnaire Response choice Coded Value Final Value Very well: could hardly 1 5 be better Pretty good 2 4 Good and bad parts 3 3 about equal Pretty bad 4 2 Very bad: could hardly 5 1 be worse The Quality of Life item is scored 1-5 before standardization. 30

31 Table 14. Scoring Information: Health Transition One item: 13 VERBATIM ITEM 13. How would you rate your physical health and emotional condition now compared to 4 weeks ago? ITEM SCORING Item 13 - This item needs to be recoded. Questionnaire Response choice Coded Value Final Value Much better 1 5 A little better 2 4 About the same 3 3 A little worse 4 2 Much worse 5 1 The Health Transition item is scored 1-5 before standardization. 31

32 Transformation of Scale Scores The final step in scale construction involves transforming the raw scale scores to a 0 to 100 scale. This transformation permits comparisons among various dimensions which may have different response categories. A score of 0 is the lowest possible score and 100 is the highest score. Formulas for linear transformations for each of the scales are as follows: Scale General Health Perception Transformation Formula Lgenheal = (100/(25-5)) * (General Health Perception raw score - 5) Physical Functioning Lphys = (100/(18-6)) * (Physical Function raw score - 6) Role Functioning Lrole = (100/ (4-2)) * (Role Function raw score - 2) Social Functioning Lsocial = (100/(6-1)) * (Social Function raw score- 1) Cognitive Functioning Lcognitiv = (100/(24-4)) * (Cognitive Function raw score -4) Pain Lpain = (100/(ll-2)) * (Pain - 2) Mental Health Lmental = (100/(30-5))*(Mental Health raw score - 5) Energy/Fatigue Lvitalit = (100/ (24-4)) *( Energy/Fatigue raw score - 4) Health Distress Ldistres = (100/(24-4)) * (Health Distress raw score - 4) Quality of Life Lquality = (100/(5-1)) * (Quality of Life raw score - 1) Health transition Ltrans = (100/ (5-1)) * (Health Transition raw score - 1) For example, using the transformation formula for the Cognitive Functioning scale, 100 = the highest possible score in the transformation; 24 = the top of the range for the sum of the untransformed item scores, while 4 = the lowest possible score of the untransformed scale. A raw score of 21 on the Cognitive Functioning scale would be transformed as follows: Lcognitiv = (100/ (24-4)) * (21-4) = (5 * 17) = 85 32