Pla d'Acció sobre Drogues de Barcelona, Institut Municipal de Salut Pública, Ajuntament de Barcelona,
1 Complejo Asistencial Benito Menni (Madrid),
2 Unitat Assistencial, Institut Català de la Salut. Direcció General Atenció Primària Mataró,
3 Hospital Universitario La Paz (Madrid) and
4 Unitat de Bioestadística, Laboratori de Medicina Computacional, Facultat de Medicina, Universitat Autònoma de Barcelona, Spain
Received 5 August 1999; in revised form 14 November 1999; accepted 28 November 1999
![]() |
ABSTRACT |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
![]() |
INTRODUCTION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Brief intervention is mainly directed toward heavy drinkers identified in an opportunistic way, when they visit the doctor for a non-alcoholic reason. The focus is on cutting down the alcohol consumption and this has frequently been shown to be effective (Poikolainen, 1999), probably attributable to personal resources underlying the change process. The process of change model, described by Prochaska and DiClemente (1986), explains this evolutionary process and permits the possibility of adjusting the therapeutic approach to the patient's stage of change, which should give better results. Having an instrument to properly allocate patients within the stages of change might be very useful for doctors working in primary health care or in general hospital wards.
Rollnick et al. (1992) proposed their self-completed Readiness to Change Questionnaire (RCQ), which seemed to have brevity and feasibility for use with brief opportunistic intervention by busy non-specialized professionals. The test considered just the pre-contemplation, contemplation and action stages, those in which a patient might be prior to or at the point of starting an intervention. Maintenance stage was not included by these authors, because of the overlapping of this stage with the pre-contemplation stage (lack of concern because of unawareness or because the problem is already overcome) and because patients at the maintenance stage would not need intervention.
The gradual introduction of brief intervention strategies in Spain justified studying the possible inclusion of RCQ in medical protocols. We contacted the authors and obtained their approval for validating a Spanish version of the questionnaire, as well as their advice during the validation process. Validating a Spanish version of the RCQ meant: (1) making a reliable translation of the test, not literally, but preserving all its original meaning; (2) administering it to a sample of patients with the aim of testing the internal consistency among scales, and the reliability between test and retest for the three scales; to determine the concurrent validity of the questionnaire against the blind external judgement; (3) ensuring that the Spanish RCQ would be giving the same information as the English one.
![]() |
SUBJECTS AND METHODS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
The AUDIT (Alcohol Use Disorders Identification Test) was developed by the WHO (Babor et al., 1989), and consists of 10 items which screen for alcohol use (items 13), alcohol-related problems (items 710) and dependence symptoms (items 46) during the past 12 months. Scoring 38 on AUDIT means at least being a hazardous or harmful drinker.
Procedure
RCQ was translated and back-translated, and the Spanish version was considered reliable and approved by the authors. Testing in 15 volunteers showed that the questionnaire was understandable.
Patients were recruited from two different health care settings: (1) a primary health care centre in Barcelona (ABS Barceloneta); (2) a general hospital ward in Madrid (Hospital La Paz). They were neither consecutive nor chosen at random. They were available to be asked to complete the AUDIT questionnaire, and were not attending for alcohol-related reasons. Illiterate and old people and supposedly alcohol-dependent patients were excluded. From each participating patient, the following information was gathered, besides informed consent: social and demographic data, alcohol consumption (current typical week) and AUDIT answers. Patients scoring 38 (seven or more in women) were included.
The selected patients were asked to fill in the RCQ. Thereafter, its result was compared with the diagnostic classification independently made by two experts (blind judgement), if possible on the very same day. Experts were four professionals (two for each sub-sample): three psychiatrists with a long clinical and research experience in alcoholism and one general practitioner trained in alcohol dependence screening and treatment. All of them were familiar with the model of change. The use of experts' blind judgement as external criteria was decided, because the procedure of using cartoons as criteria where patients had to tick one of four cartoons representing the four selected stages proved erratic.
Two experts (A and B) in each setting (primary health centre and hospital ward) separately and consecutively allocated each subject to a stage of change according to their professional opinion after a clinical interview (~20 min) with the patient. They had no previous information either on the patients or on the test score. They did not intervene either in patients' further assessment or advice: any appointment to tackle the drinking pattern and its consequences was postponed and done by other professionals.
Dependent variables were raw scores obtained for each scale of the RCQ. The hypothesis was that subjects would score highest in the scale to which the experts had allocated them.
The RCQ was given to all 201 patients. A second retest was performed 2 days later to a reduced number of patients (testretest). This short period between measurement points was recommended by the authors.
To complete the sample was not easy, especially in the primary care setting, because of patients' reluctance to attend appointments, even if paid. Another difficulty was the low educational level, particularly in outpatients (the centre belonged to a harbour neighbourhood), some of them being excluded because of illiteracy.
Data analysis
Statistical analysis was carried out using the statistical package SPSS. Demographic characteristics of the sample are presented as descriptive statistics (percentage, mean ± SD). Comparisons of means were made using Student's t-test and analysis of variance. The 2-test was used for comparing percentages. A 5% significance level was accepted for all the tests. Cronbach's alpha coefficient, factor analysis of principal components by correlation with Varimax rotation, and weighted kappa coefficient for testretest were calculated to analyse the reliability of RCQ.
![]() |
RESULTS AND DISCUSSION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Data on age, sex, work category and education level for patients at both sub-samples showed no significant differences between centres. Accordingly, data from the two centres were combined (Table 1).There was a predominance of male patients, active, manual workers of average educational achievements and a mean (± SD) age of 46.0 ± 10.7 years.
|
Reliability analysis showed the following mean scores on each scale: pre-contemplation scale: 0.98 (± 3.46); contemplation scale: 0.75 (± 4.18); action scale: 0.29 (± 4.50).
Cronbach's alpha coefficient was calculated for the four items composing scale representing a stage of change. Within each scale, there was a positive relationship between every item, so that item scores can be regarded as constituting a scale: pre-contemplation: alpha 0.58 (0.73, in the English version); contemplation: alpha 0.75 (0.80, in the English version); action scale: alpha 0.80 (0.85, in the English version).
Factor analysis. Factor and conglomerate analysis (Table 2) gave the same type of results, showing three clearly identified components.
|
Factor analysis of the principal components of the questionnaire with Varimax rotation showed a clear factor structure corresponding to the three stages of change. When considered together, the three factors account for 57.4% of the total variance.
This internal consistency can also be observed on the graphical representation (see Fig. 1) of components in rotation space, where there appear three factors accounting each for: (1) action, 23.6% of the variance; (2) contemplation, 18.3% of the variance; (3) pre-contemplation, 15.6% of the variance.
|
Allocation to stage of change
Allocation of patients to one of the stages of change was based on the possible different response patterns, according to its positive or negative sign on each of the three scales (pre-contemplation; contemplation or action). Definitive allocation of subjects to each stage was established according to the highest raw score obtained among the three scales (disregarding its positive or negative value). In the case of a tie between two adjacent scale scores, the subject was allocated to the stage farther along the continuum, following the authors' recommendation to assume that the patient had reached the furthest point in the change process. Theoretically, no ties should occur between opposite scales (pre-contemplation and action) or between all three scales. In this Spanish study, 24 ties were detected (including six ties between pre-contemplation and action). Thus, the 201 subjects were allocated as follows: 81 on action scale (40.3%); 78 on contemplation scale (38.8%); 42 on pre-contemplation scale (20.9%).
The pre-contemplation scale showed a negative correlation with contemplation and action scales. As regards concurrent validation, agreement with the RCQ assessed as weighted kappa was 0.44 with expert A and 0.52 with expert B. Between A and B experts weighted kappa was 0.92 (n = 201).
Variables' influence on the questionnaire scales
The selected variables (AUDIT score, alcohol consumption, sex and expert's judgement) were crossed with the following result (Table 3 shows the AUDIT mean score for each allocation stage). Analysis of variance for the three groups showed significant main effects (F = 8.6, df = 2.198; P < 0.001) and, using post hoc Dunnett's test C, statistically significant differences between contemplation and action. Scores were coherent with the foreseen target population: heavy drinkers, with or without added problems but without suspected dependence.
|
Contemplation and action scales were different concerning alcohol consumption based on Scheffé's a posteriori test, with a higher consumption in those on contemplation.
Among patients with a tie in scores belonging to different stages, a PC (pre-contemplation = contemplation) tie was observed mainly in patients with a high alcohol consumption (83%). Of subjects with a PA (pre-contemplation = action) tie, 66.7% show a low alcohol intake. Subjects with CA (contemplation = action) tie showed either low, medium or high consumption. Of PCA ties (pre-contemplation = contemplation = action), 66.7% had a low consumption.
No significant differences were observed concerning total RCQ scores and sex. When both sexes are separately considered, stage of change calculated by RCQ compared to expert's blind external judgement revealed that agreement was greater in women (weighted kappa 0.56) than in men (weighted kappa 0.41).
In the Spanish version of the RCQ, the three loading factors account for 57.4% of the total variance, a somewhat smaller percentage than the Australian sample, where the first three factors together account for 68.6% of total variance: 46.1% for the action component; 12.6% for the contemplation component, and 9.9% for the pre-contemplation component. Nevertheless, when comparing explained variance for each scale between the original version and the Spanish one, we observed that variance was higher in the English version (23.6% compared to 46.1%) only for the action scale, whereas explained variances for pre-contemplation and contemplation scales were superior in the Spanish sample (15.6% vs 9.9%, 18.3% vs 12.6%, respectively).
Internal consistency was observed between items belonging to the same scale. Relationship between adjacent scales (contemplation and action) was significantly greater than between non-adjacent scales, as was the case in the original version.
Though adequate, Cronbach values were smaller than those obtained by the original authors, especially for the pre-contemplation scale (0.58), with better results with progressing stage of change (contemplation and action). The concurrent validity, when facing the instrument with external validation criteria, gave a limited kappa value (0.44 with expert A and 0.52 for expert B).
Understanding the questionnaire is not easy, especially when it deals with double negative formulations. Nevertheless, rewording the items would have changed the original instrument.
Discrepancies between test and expert judgement were highest in those patients positively scoring in more than one scale, especially when these were adjacent (contemplationaction).
Ties and discrepancies between RCQ classification and external experts' judgement made us consider response patterns also, a possibility not previously foreseen, because of the opinion of the original authors that raw scores were more operative. Anyway, discrepancies between both ways of evaluating the test (total scores and score patterns) were not as noteworthy as those between test and experts' blind judgement. We therefore considered that it did not make sense to use the refined method (to diagnose through patterns), but, instead, to use the global score of the short RCQ. The authors of the English version reached a similar conclusion.
In the light of the limited concurrent validity of RCQ in this study, we considered four possible confounding factors:
(1) The test's structure itself, with three possible scales where readiness to change is indistinguishable from action and there is no maintenance scale, may have contributed to a wrong allocation of those patients who didn't find their place. The lack of a delimited readiness to change or preparation stage seems not to be a problem. This is probably due to its position between the adjacent scales of contemplation and action, which follow each other without any gap and also because of solving any tie by allocating the patient in the most advanced stage.
Patients in the maintenance stage might have been more problematic. Some of them had been included in the sample, because the AUDIT was used as our screening tool. AUDIT is capable of detecting alcohol-related problems occurring in the last 12 months among patients who are now light drinkers or not drinking any longer. These patients can both have been located in the action scale or in the pre-contemplation scale, depending on their current attitude (believing that they are still doing something concerning their drinking or perhaps considering that they have no problems and have already turned a page). To reframe the questionnaire by adding a fourth scale for maintenance goes beyond the aim of a validation study. Because the test is supposed to have been designed for pointing out the most suitable intervention after an opportunistic detection, theoretically RCQ would not be given to patients in the maintenance stage.
(2) Abstainers or patients with low alcohol consumption (38.3% of the sample) might be responsible for the test's poor concurrent validity. To control for this possible confounding factor, the principal component analysis was repeated for a sub-sample made up of only those patients of a moderate to high risk consumption (>21 units/week and >35 units/week, in women and men respectively). This did not improve the global accumulated variance (57.5), each factor accounting respectively for the following percentage variance: action: 25.2%; contemplation: 17.1%; pre-contemplation, 15.2% (Table 4).
|
|
|
![]() |
ACKNOWLEDGEMENTS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
![]() |
FOOTNOTES |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
![]() |
REFERENCES |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Heather, N., Rollnick, S. and Bell, A. (1993) Predictive validity of the Readiness to Change Questionnaire. Addiction 88, 16671677.[ISI][Medline]
Poikolainen, K. (1999) Effectiveness of brief interventions to reduce alcohol intake in primary health care populations: a meta-analysis. Preventive Medicine 28, 503509.[ISI][Medline]
Prochaska, J. O. and DiClemente, C. C. (1986) Toward a comprehensive model of change. In Treating Addictive Behaviors: Processes of Change, Miller, W.R. and Heather, N. eds. Plenum Press, New York.
Rollnick, S., Heather, N., Gold, R. and Hall, W. (1992) Development of a short Readiness to Change Questionnaire for use in brief, opportunistic interventions among excessive drinkers. British Journal of Addiction 87, 743754.[ISI][Medline]