1.2 JUSTIFICACIÓN DE INVESTIGACIÓN
2.2.9 TIPOS DE DATA CLUSTERING
A review of the methodology resulted in several changes in the QOLAS between Study 1 and Study 2. A copy of the revised QOLAS can be xound in Appendix 5).
First, the rating method was simplified by changing from a 10-item dashed analogue scale to a 5-item rating scale ranging from 'Not a problem' to 'It could not be worse'.
Secondly, three of the elements used in the original version (AGE 25, AGE 50 and AGE 75) were changed to 5 YEARS AGO, IN 5 YEARS TIME, IN 10 YEARS TIME. These were then subsequently revised to MOTHER, FATHER, AS OTHERS SEE YOU. These changes were made as it was found that patients had difficulties in making judgements about hypothetical situations (for example, how do you think you will be at the age of 75 or in 10 years time) . On the premise that quality of life is a comparative phenomena in which a person's view of their own life situation in relation to important others (for example, family members and peers) is a salient factor in
their overall judgement of quality of life, it was considered that the elements (MOTHER, FATHER, AS OTHERS SEE YOU) might be more appropriate. The remaining seven elements (NOW, BEFORE,
EXPECT, LIKE, FRIEND, BEST, WORST) were unchanged between studies.
Thirdly, the number of constructs elicited was increased from ten to fifteen. During the construct elicitation procedure it was found that in several instances a patient would spontaneously report many areas of importance in one specific life domain, for example social functioning, while none were forthcoming in other domains. By restricting each patient to two items within each domain and forcing them to choose two within a domain that may not have been of particular relevance to them, it was felt that items of importance may be excluded. In addition, constructs were sometimes elicited which did not readily fit into any of the five domains considered (physical, cognitive, emotional, social, economic/employment), for example, having faith in God. Thus, in Study 2, five free constructs were added to the questionnaire, allowing patients more freedom in the items of importance chosen. Full details of the changes made during the development of the test instrument can be found in Table 4.5.
4 . 3 . C . Ü Sickness Impact Profile (SIP) (Bergner et al., 1981) This is a general health status questionnaire containing 1T6 statements about health-related dysfunction and covering 12 areas of functioning: ambulation, bodycare and movement, alertness behaviour, social interaction, emotional behaviour, communication, sleep and rest, eating, work, household management and recreation/pastimes. Patients are asked to read the statements and endorse or tick only those statements that describe them at present and are related to their health. A scaled score is assigned to each item, based on the judgement of 25 'experts' who rated each SIP item on a 15-point scale of dysfunction. Percentage disability scores for each category are calculated by summing the scale values of all items checked in that category, dividing this sum by the maximum possible score for that category and multiplying the obtained quotient by 100. A score of 100 indicates maximal dysfunction.
Table 4.5: Details of the development of the Quality of Life Assessment Schedule in patients with epilepsy.
Study Patient numbers Number of constructs Rating method Elements 1 1-10 10 100 VAS AGE 25 AGE 50 AGE 75 11-24 10 1-10 VAS AGE 25 AGE 50 AGE 75 2 25-31 10 1-5 LIKERT -5 YEARS +5 YEARS +10 YEAR 32-50 15 1-5 LIKERT MOTHER FATHER OTHERS
In addition, three composite scores can be computed, on the same basis, which assess physical, psychosocial and overall functioning. A copy of the SIP can be found in Appendix 6. Psychometric properties of the SIP
There are many reports in the literature concerning the reliability and validity of this measure (Carter et a l ., 1976; Pollard et al., 1976; Bergner et al., 1976; Bergner et al., 1981). Work began on this questionnaire in 1972. The first version contained 312 items, based on statements derived from interviews with patients, carers, people without illness and health care professionals. Subsequent item analysis, on data obtained from 246 individuals (including patients and non patients) , resulted in a reduction from 312 to 189 items. The 189-item SIP was subjected to rigorous testing in trials conducted in 1974 and 1976, with particular attention being paid to the issues of reliability and validity. A conscious attempt was made to test as broad a range as possible of subjects with health-related dysfunction. Samples of subjects assessed have included: GP enrollees, rehabilitation out and in-patients, speech pathology inpatients, outpatients with chronic health problems, hyperthyroid patients, patients with rheumatoid arthritis and patients who had undergone hip replacements.
High test-retest stability coefficients have been noted in both trials. In the 1974 trial, 199 subjects completed the SIP questionnaire on 2 occasions with a test-retest interval of 24 hours. Two forms of the questionnaire: long version (235 items) vs short version (146 items) and 2 types of administration: interviewer and self-administered were compared. 6 interviewers were used. The test-retest reliability coefficients of category scores ranged between 0.75 and 0.92, dependent upon type of administration, different interviewers and severity of dysfunction of the patient group assessed. Reliability in terms of whether or not the same items were checked was lower (r=0.45-0.60). In the 1976 trial, 24-hour test retest reliability (n=53) for category and overall scores was high (r=0.92), however the reproducibility of items checked was lower (r=0.50).
Reliability was highest for self-administered, interviewer supervised mode of administration (r=0.97), with interviewer administration producing an lower reliability (r=0.87). No data were available regarding the test-retest reliability of questionnaires completed postally. However, the authors feel that postal questionnaires provide less reliable information as the correlations between the postally completed SIP scores and external criterion (for example, self assessment of dysfunction) were lower than for other types of administration. Internal consistency, as assessed by Cronbachs alpha, was also high (r=0.94).
The 1974 trial provided some evidence of the validity of the SIP in that it successfully discriminated between sub groups of patients with varying degrees of dysfunction. In addition, SIP scores correlated with external criterion measures (patient and clinician subjective assessments of dysfunction). In 1976, a more detailed analysis of the convergent and discriminant validity of the SIP was performed utilising the multitrait-multimethod approach (Campbell and Fiske, 1959). In this analysis, 5 external criteria were assessed: 1) self assessment of dysfunction (SAD), 2) self assessment of sickness (SAS), 3) National Health Survey Index of Activity Limitation, Work Loss and Bed Days (NHIS) , 4) clinician rating of dysfunction (CAD) and 5) clinician rating of sickness (CAS). Several hypotheses regarding the relationships between these variables were proposed and a hierarchy of correlations suggested: SIP and SAD > SIP and SAS > SIP and NHIS > SIP and CAD > SIP and CAS. This hierarchy was confirmed by the analysis providing some evidence of convergent and discriminant validity. In addition, the SIP category scores had a high mean test-retest correlation (r=0.82, SD=0.08) with a low mean correlation being seen between categories (r=0.32, SD=0.19 (Time 1); r=0.40, SD=0.21
(Time 2).
A further review of the content of the 189-item SIP resulted in the final 13 6-item version. The data from the 189- item version were re-scored (based on 136-items only) and re analysed. The 136-item version performed as well or better than the longer version in terms of reliability and validity.
4.3.c.iii Mood Adjective Checklist (MACL) (McNair and Lorr, 1964; Lishman, 1972)
The Lishman version of the Mood Adjective Checklist (McNair and Lorr, 1964) was used to assess mood at time of testing. This 24 item checklist provides objective measures of anxiety, hostility, vigour, depression and fatigue. Patients are asked to rate how they feel on a 4-point scale, ranging from 'not at all' to 'extremely', on each of the 24 adjectives listed (see Appendix 7 for copy of this scale).
Psvchometric properties of MACL
The version used in this study was of the same format as that used by Lishman (1972) in a study of the affective components of memory functioning. The study was conducted on psychiatric inpatients suffering from affective disorder. In this study, 3 measures of depression were employed: a clinical rating of depression; the Beck Depression Inventory and the depression sub-scale of the MACL. The author reports a high correlation between the MACL Depression score and the Beck Depression Inventory (r=0.96, p<.05) and a moderate correlation between the MACL Depression score and the clinical rating of depression (r=0.64, p < 0 5 ) , thus providing evidence of concurrent validity for this sub-scale. No other information regarding the other sub-scales of the MACL is given.
The version used was adapted from that described by McNair and Lorr (1964), originally containing 55 adjectives, covering five hypothesised moods: tension, anger, depression, vigour and fatigue. The factorial and concurrent validity of the mood factors and their sensitivity were determined by three main studies in which a total of 853 male psychiatric outpatients and 45 normal controls were assessed. In two of the studies the MACL was administered during treatment, while in the third the assessment was a one-off assessment. The form of the MACL varied across the three studies, reducing from 55
items in Study 1 to 38 items in Study 3.
The five factors proposed were confirmed by factor analysis of the items. In addition, two further factors relating to Confusion and Friendliness also emerged, though these were weaker factors. In addition, fatigue and vigour.
while negatively correlated do appear to be separate factors and not opposite poles of a bipolar factor. The five mood factors showed high congruence across the three studies, ie. the same factors were apparent in three separate studies, suggesting high validity of the proposed factors.
Evidence of test-retest reliability is reported from one study in which 150 male psychiatric outpatients were assessed prior to and following 4 weeks of treatment. The MACL administered in this study consisted of 38 adjectives covering 6 mood states: tension (8 items); anger (5 items); depression (9 items); vigour (6 items), fatigue (5 items) and confusion (5 item s ) . Moderate correlation coefficients were seen for all scales (tension, r=0.64; anger, r=0.68; depression, r=0.69; vigour, r=0.64; fatigue, r=0.61; confusion, r=0.61).
Some evidence of the sensitivity of this measure is provided by one study (Study 3) in which 180 male psychiatric patients were compared to 45 normal controls. Significant differences were seen on all scales at the .001 level, with patients reporting higher levels of tension, anxiety, depression and fatigue and lower levels of vigour than controls. In the same study, in which patients were being treated with either drug therapy, psychotherapy or placebo, the MACL successfully detected drug/placebo differences in the experimental group, while the controls showed no evidence of mood change. This suggests that the questionnaire is sensitive to treatment effects. Work by Thompson and Trimble (1982b) in patients with epilepsy also demonstrates the sensitivity of the MACL in detecting drug-related changes. In comparison to a control group, who did not undergo any change to their antiepileptic medication, patients who were undergoing a drug reduction (from a mean of 2.8 drugs to a mean of 1.6 drugs) demonstrated a significant fall in anxiety as assessed by the MACL. This change in anxiety was also seen on another questionnaire, the Middlesex Hospital Questionnaire (MHQ)
(Crown and Crisp, 1966). In addition, a group of patients whose therapy was changed by substituting carbamazepine therapy for one or more drugs, showed a significant reduction in anxiety and an increase in vigour. These changes, however, were not seen on the MHQ.
Scanty evidence regarding concurrent validity is given. Low, but significant correlations have been shown between MACL scales and independent measures of interpersonal behaviour (Interpersonal Behaviour Inventory, Lorr and McNair, 1963) and anxiety (Taylors's (1953) Manifest Anxiety scale) (McNair and Lorr, 1964).
The MACL appears to be minimally sensitive to social desirability of response set with 4 of the 5 scales showing low to moderate correlations with a scale of social desirability (tension, r=-.21; depression, r=-.36; vigour, r=.33; fatigue, r=-.18). However, the factor 'anger' did show a moderate correlation with social desirability (r=-.52).
In summary, data relating to the form of the MACL used in this study is scanty and chaotic, the reported reliability and sensitivity data arising from a number of studies with varying numbers of adjectives employed. No clear structure seems apparent for the testing of the psychometric properties of this test, the pervading feeling being that the analyses were performed ad hoc. Regarding test-retest reliability, it is difficult to equate this with a correlation between scores obtained on two separate occasions in which a change in treatment has occurred in the intervening interval.
4.3.c.iv Life Events Schedule (LES) (Sarason et al., 1978)