| | A large-scale validation study of the Medication Adherence Rating Scale (MARS)Received 28 March 2007; received in revised form 8 October 2007; accepted 29 October 2007. published online 17 December 2007. Abstract Adherence to medication is an important predictor of illness course and outcome in psychosis. The Medication Adherence Rating Scale (MARS) is a ten-item self-report measure of medication adherence in psychosis [Thompson, K., Kulkarni, J., Sergejew, A.A., 2000. Reliability and validity of a new Medication Adherence Rating Scale (MARS) for the psychoses. Schizophrenia Research. 42. 241–247]. Although initial results suggested that the scale has good reliability and validity, the development sample was small. The current study aimed to establish the psychometric properties of the MARS in a sample over four times larger. The scale was administered to 277 individuals with psychosis, along with measures of insight and psychopathology. Medication adherence was independently rated by each individual's keyworker. Results showed the internal consistency of the MARS to be lower than in the original sample, though adequate. MARS total score correlated weakly with keyworker-rated adherence, hence concurrent validity of the scale appeared only moderate to weak. The three factor structure of the MARS was replicated. Examination of the factor scores suggested that the factor 1 total score, which corresponds to the Medication Adherence Questionnaire [Morisky,D.E., Green,L.W. and Levine,D.M., 1986. Concurrent and predictive validity of a self-reported measure of medication adherence. Medical Care. 24, 67–74] may be a preferable measure of medication adherence behaviour to the total scale score. 1. Introduction  Medication adherence behaviour lies on a continuum from complete adherence to prescribed medication, through partial adherence, to complete non-adherence. Non-adherence to prescribed antipsychotic medication is associated with symptomatic relapse (Ayuso-Gutierrez and del Rio Vega, 1997, Fenton et al., 1997, Robinson et al., 1999, Chen et al., 2005), hospital admission (Haywood et al., 1995, Gray et al., 2002, Leucht and Heres, 2006), and poor outcome (Helgason, 1990, Olfson et al., 2000, Leucht and Heres, 2006). Estimates of the frequency of non-adherence to antipsychotic medication vary, with review studies suggesting the rate lies between 25 and 55% (Young et al., 1986, Fenton et al., 1997, Nosé et al., 2003). The problem of non-adherence is therefore substantial. Finding ways of reliably determining whether individuals are following treatment recommendations, and of improving rates of adherence among those who are not, are clearly pertinent issues in clinical practice. There are a number of approaches to studying medication-taking behaviour. The main methods are patient and clinician report, pill counts, and biological methods (such as blood or urine drug concentrations), each of which confers different advantages and disadvantages (Farmer, 1999). Self-report measures have the benefits of being cheap, easy to administer, non-intrusive, and able to provide information on attitudes and beliefs about medication. Potential limitations to self-report are that ability to understand the items, and willingness to disclose information, can affect response accuracy and thus questionnaire validity. The Medication Adherence Rating Scale (MARS, Thompson et al., 2000) is a ten-item yes/no self-report instrument. It was developed from two existing scales, the 30-item Drug Attitudes Inventory (DAI; Hogan et al., 1983) and the 4-item Medication Adherence Questionnaire (MAQ; Morisky et al., 1986), with the aim of developing a more reliable and valid tool for assessing medication adherence behaviour in psychosis. Total scores range from 0 (low likelihood of medication adherence) to 10 (high likelihood). This reflects an understanding that adherence is a continuous variable: An individual can reach a decision anywhere between complete adherence and complete non-adherence, such as only taking medication when they feel unwell (see Perkins, 1999, Perkins, 2002). The development sample for the MARS consisted of 66 participants with psychosis and the scale showed good internal consistency (alpha =.75). Three factors were identified. These were considered to represent ‘medication adherence behaviour’ (items 1–4), ‘attitude toward taking medication’ (items 5–8) and ‘negative side-effects and attitudes to psychotropic medication’ (items 9,10). A positive correlation of 0.6 between the MARS adherence score and blood lithium levels suggested construct validity. The sample size for this analysis was small, however (n=17), and as the authors acknowledge, the participants' awareness that a blood sample would be taken may have altered the results. The MARS total score did not correlate with carer-rated medication adherence, which Thompson et al explain by reference to studies questioning the accuracy of carer-report (Cochran and Gitlin, 1988, Piatkowska and Farnhill, 1992, Fenton et al., 1997). While this may indeed be the case, the result may still reflect a weakness in the MARS, thus increasing the need for further validation of the measure. In addition, the MARS has yet to be examined for its ability to reproduce established relationships in the medication adherence literature. Two consistent findings are that good medication adherence is related to greater insight into illness and less psychopathology (Nosé et al., 2003, Kampman and Lehtinen, 1999, Fenton et al., 1997, Mcevoy et al., 2006). Insight is a dimensional phenomena (David, 1990), and the dimension shown to be most implicated in treatment adherence is insight into need for treatment (Buckley et al., 2007). The MARS was selected as a measure of medication adherence in the Psychological Prevention of Relapse in Psychosis (PRP) trial, a randomised controlled trial of cognitive behavioural therapy for the positive symptoms of psychosis with 301 participants. The size of the sample makes this an ideal opportunity to re-examine the psychometric properties of the MARS. The aims of the current study are therefore to provide reference data for the MARS, to examine its reliability and validity in a large sample of individuals with psychosis, to re-examine its factor structure, and to investigate its relationship with established correlates of medication adherence. 2. Method 2.1. Participants Two-hundred and seventy-seven participants were included, from the 301 patients who took part in the Psychological Prevention of Relapse in Psychosis (PRP) Trial (ISRCTN83557988). In order to be eligible for the PRP trial, participants were required to have relapsed not more than 3 months before consent was obtained. They were excluded if they had a primary diagnosis of alcohol or substance dependency, organic syndrome, or learning disability, or a command of English insufficient to engage in psychological therapy. In recruiting for the PRP trial, a total of 683 patients meeting inclusion criteria were identified, of whom 301 patients (44%) provided informed consent and 382 patients withheld consent to the trial. Those who consented did not differ in age from those who did not, but they were more likely to be male (chi square 8.23, df 1, p=0.004). The 277 individuals included in this current analysis comprised those who met the additional inclusion criteria of being been prescribed antipsychotic medication and having completed the MARS. Eighty-two (29.6%) were female and 195 (70.4%) male. All patients met criteria (ICD-10, F20) for schizophrenia (84.5%), schizoaffective disorder (14.4%) or delusional disorder (1.1%). Their mean age was 37.6 years (S.D.=11.09) and the mean duration of illness was 11.0 years (S.D.=9.0). One hundred and eighty three (66.1%) were admitted to hospital during their relapse, and 94 (33.9%) were outpatients throughout. The mean PANSS positive symptom score was 18.1 (SD=5.2), the mean PANSS negative symptom score was 13.1 (SD=5.7) and the mean PANSS general psychopathology score was 33.6 (SD=7.6). For comparison, the mean scores from 101 people from a long-term psychiatric unit reported by Kay et al. (1987) in their development of the PANSS were: PANSS Positive=18.2 (SD=6.1), PANSS Negative=21.0 (SD=6.2), and PANSS General Psychopathology=37.7 (SD=9.5). Thus the current relapsing sample had the same level of positive symptoms but lower levels of negative symptoms compared with a chronic group. Patients reported their ethnicity as White (71.82%), Black-African (9.0%), Black-Caribbean (7.6%), Black-other (2.5%), Indian (1.8%) and Other (7.2%). Most were unemployed (79.1%), with the remainder being full-time employed (5.4%), part-time employed (2.5%), voluntarily employed (3.6%) or other (9.4%). A minority of the sample was married or in a stable relationship (10.8%), the remainder being single (72.2%), separated/divorced (14.8%) or widowed/unknown (1.5%). All participants included in the study were taking antipsychotic medication. Most (81.9%) were taking a single antipsychotic preparation but some were prescribed two (17.3%) or three (0.7%) antipsychotic drugs. Fifty-seven participants (20.6%) were also taking antidepressant medication. Overall doses of antipsychotic medication were low (equivalent to 0–200 mg chlorpromazine; 33.2%), medium (equivalent to 200–400 mg chlorpromazine; 41.5%), high (equivalent to more than 400 mg Chlorpromazine; 24.2%) and dose unknown (1.1%). 2.2. Recruitment The PRP trial is a UK multi-centre randomised controlled trial of cognitive behaviour therapy and family intervention for psychosis. Recruitment was from specified clinical teams in London and East Anglia, with the aim of obtaining a representative sample of individuals with psychosis. Patients meeting the eligibility criteria were asked to provide informed consent for participation by a trial research worker or research clinical psychologist. 2.3. Measures The assessment was completed by graduate research psychologists after patient consent had been obtained, and before randomisation. During the assessment, patients were asked to report the type and dose of medication they were taking. In addition the following measures were used: Medication Adherence Rating Scale (MARS; Thompson et al., 2000), described above, and items shown in Table 1. Participants in hospital during the assessment phase were asked to complete the behaviour component of the questionnaire based on the time immediately prior to their admission. Participants were aware that their answers to the MARS were confidential to the trial and would not be shared with their clinical team. Engagement measure (Hall et al., 2001). An 11-item instrument designed to measure client engagement with mental health services. Clients were rated by mental health professionals who were unaware of MARS score. High scores indicate greater levels of engagement. Item 6 measures medication compliance: “Extent to which client agrees to take medication and will take it freely”, rated from never (1) to always (5). Test–retest reliability (.76) and inter-rater reliability (.93) are good for this item. Positive and Negative Syndrome Scale (PANSS; Kay et al., 1987). This well known instrument assesses phenomena associated with schizophrenia. It provides a total score and sub-scores for ‘positive’, ‘negative’, and ‘general psychopathology’ symptoms. Higher scores indicate more severe symptoms. Symptoms are rated over the last 72 h. Assessors were trained to a high level of inter-rater reliability: for 53 PANSS positive symptom assessments rated by at least one other assessor the mean one-way random model single measure Intraclass Correlation Coefficient for pairs of raters was .88. Insight-Scale of Unawareness of Mental Disorder (SUMD: Amador et al., 1993). The SUMD is based on a multidimensional view of insight. Component dimensions (awareness of mental disorder, awareness of the achieved effects of medication, awareness of the social consequences of mental disorder, awareness of hallucinations and awareness of delusions) are rated on a five-point scale (1-aware to 5-unaware) on the basis of direct patient interview. 2.4. Statistical methods Analyses were conducted using SPSS for Windows (version 12.0.1). Parametric tests were performed where possible, but some variables, including the MARS factor total scores, did not meet the criteria for parametric testing, and non-parametric test results are reported as appropriate. For the assessment of test–retest reliability, Pearson's r is reported. Significance tests are quoted as two-tailed probabilities. 3. Results 3.2. Reliability and factor structure The internal consistency of the MARS (using Cronbach's alpha, which is equivalent to Kuder–Richardson 20 coefficient, for binary data) was 0.60. Alpha was reduced or unchanged if any items were deleted from the scale, suggesting that there are no redundant items. This moderate to low alpha value indicates that the scale may not be unidimensional, and the factor structure of the MARS was accordingly investigated. A principal components factor analysis with varimax rotation, retaining factors with an Eigenvalue greater than 1, produced a three-factor solution (Table 2). After rotation, factor 1 accounted for 20.7% of the variance, factor 2 for 15.9% and factor 3 for 13.8%. Thus, in total the rotated factor solution accounted for 50.5% of the total variance. This is the same three-factor solution as produced by Thompson et al., in which they considered the factors to represent ‘medication adherence behaviour’, ‘attitude to taking medication’ and ‘negative side-effects and attitudes to psychotropic medication’. The internal consistency of the three factors was examined and factor 1 was found to be the most internally consistent, with Cronbach's alpha of .67. Factor 2 had Cronbach's alpha of .44, and factor 3 (which only consisted of two items) had a Cronbach's alpha value of .53. 4. Discussion The full range of MARS total scores (0–10) was present in this sample of individuals with a recent relapse in positive symptoms of psychosis, indicating widely varying attitudes towards taking medication. MARS total score correlated with keyworker-rated medication adherence, but at lower than r=0.20, this represents a weak relationship, accounting for less than 4% of the variance. This suggests that the evidence produced by Thompson et al. (2000) for the validity of the MARS total score as a measure of medication adherence behaviour should be treated with some caution. Other results are suggestive of validity of the MARS. Individuals prescribed higher levels of medication were found to score more highly on the MARS. Interpretation of finding is not clear, but it may be that individuals identified by their doctors as adherent are prescribed more medication, in which case the MARS total score is shown to discriminate an adherent group. The MARS total score also reproduced the expected relationships of higher adherence with more insight into need for medication, and higher adherence with less psychopathology. However, once again correlations were weak. Our results show that the MARS is reliable even over a long interval of twelve months. Analysis over a period of a few days or weeks would be likely to further increase the test–retest reliability of the MARS. The internal consistency of the MARS was moderate (alpha=.60), but lower than the value produced by Thompson et al. (2000) during the original development of the scale (alpha=.75). This may not represent a weakness of the scale, however, as there are reasons to expect a reduced alpha value for scales with the format of the MARS, notably the binary response choice, small number of items, and scale multidimensionality. Though it is likely that the internal consistency of the MARS could be improved either by adding more response options or by adding more items, it is debatable whether this would constitute an improvement to the measure, or whether it would compromise its quick, simple format. In addition to replicating the three factors of the MARS, the current study examined the relationships of the individual factor scores. Factor 1 (adherence behaviour) was the only factor total score to correlate with keyworker-rated medication adherence. This factor correlation was of equal strength to the whole scale correlation. It therefore appears that where the concern is simply whether or not someone is taking their medication, factor 1 may be a better indicator than the whole MARS scale. This four item subscale (which corresponds to Morisky et al's Medication Adherence Questionnaire (MAQ), from which the MARS was developed) is quicker to administer, has a higher internal consistency than the overall scale, and appears valid for this purpose. Factor 2 total score (attitude towards medication) correlated with insight into illness and insight into the effects of medication but failed to correlate with keyworker rated adherence. This finding that attitude does not translate into adherence behaviour is not surprising when the multiple factors involved in determining adherence behaviour are considered (Perkins, 1999, Perkins, 2002). For example, someone with a positive attitude towards medication may forget to take it, or someone with a negative attitude may take medication to please someone else. The current study has a number of limitations. A single item keyworker rating for medication adherence was used to determine the concurrent validity of the MARS. Although the proportions of participants rated as being adherent (69%) or not adherent (31%) to their medication regimen using this method was in line with previous estimates (Young et al., 1986, Fenton et al., 1997, Nosé et al., 2003), a multi-item measure would have been preferable. The sample was mixed, containing both inpatients and outpatients, and it is possible that inaccuracy was introduced due to retrospective completion among inpatients. The results do not identify an effect of inpatient completion of the measure, but placing this extra memory demand on questionnaire completion, especially among a group who may have cognitive deficits, was not ideal. An all-outpatient sample would have been preferable to eliminate this problem. Furthermore, all participants had consented to a treatment trial, so may not be fully representative of those with psychosis. In particular, they are likely to be a better engaged group. To summarise, the MARS is a quick, non-intrusive measure of medication adherence. Its reliability is adequate, but validity appears only moderate-weak. Items in the MARS about attitude to medication may be informative to clinicians identifying barriers to adherence in individual cases, but do not appear to be valuable in predicting adherence behaviour over a large sample. Factor 1 (medication adherence behaviour), corresponding to the Medication Adherence Questionnaire (MAQ), may be superior for this purpose. Further demonstrations of test–retest reliability and validity of the MARS are recommended. Role of funding source Wellcome Trust. Contributors Laura Fialko Department of Psychology, Institute of Psychiatry, King's College London, University of London, UK. Philippa A. Garety Department of Psychology, Institute of Psychiatry, King's College London, University of London, UK. Elizabeth Kuipers Department of Psychology, Institute of Psychiatry, King's College London, University of London, UK. Graham Dunn Biostatistics Group, School of Epidemiology & Health Sciences, University of Manchester, UK. Paul E. Bebbington Department of Psychiatry and Behavioural Sciences, Royal Free and University College Medical School, University College London, University of London, UK. David Fowler School of Medicine, Health Policy and Practice, University of East Anglia, UK. Daniel Freeman Department of Psychology, Institute of Psychiatry, King's College London, University of London, UK. Conflict of interest None. Acknowledgement This work was supported by a programme grant from the Wellcome Trust (No. 062452). We wish to thank the patients taking part in the trial and the participating clinical teams in the four NHS Trusts. The study was supported by Camden and Islington Mental Health and Social Care Trust, North East London Mental Health NHS Trust, Norfolk and Waveney Mental Health Partnership NHS Trust and South London and Maudsley NHS Trust. References Amador et al., 1993. 1.Amador XF, Strauss DH, Yale SA, Flaum MM, Endicott J, Gorman JM. Assessment of insight in psychosis. American Journal of Psychiatry. 1993;150(6):873–879. Ayuso-Gutierrez and del Rio Vega, 1997. 2.Ayuso-Gutierrez JL, del Rio Vega JM. Factors influencing relapse in the long-term course of schizophrenia. Schizophrenia Research. 1997;28:199–206. Abstract |
Full-Text PDF (624 KB)
|
CrossRef
Buckley et al., 2007. 3.Buckley PF, Wirshing DA, Bhushan P, Pierre JM, Resnick SA, Wirshing WC. Lack of insight in schizophrenia: impact on treatment adherence. CNS Drugs. 2007;21:129–141. MEDLINE |
CrossRef
Chen et al., 2005. 4.Chen EYH, Hui CLM, Dunn ELW, Miao MYK, Yeung WS, Wong CK, et al. A prospective 3-year longitudinal study of cognitive predictors of relapse in first-episode schizophrenic patients. Schizophrenia Research. 2005;77:99–104. Abstract | Full Text |
Full-Text PDF (95 KB)
|
CrossRef
Cochran and Gitlin, 1988. 5.Cochran SD, Gitlin MJ. Attitudinal correlates of lithium compliance in bipolar affective disorders. Journal of Nervous and Mental Disease. 1988;176:457–464. MEDLINE |
CrossRef
David, 1990. 6.David AS. Insight and psychosis. British Journal of Psychiatry. 1990;156:798–808. MEDLINE |
CrossRef
Farmer, 1999. 7.Farmer KC. Methods for measuring and monitoring medication regimen adherence in clinical trials and clinical practice. Clinical Therapeutics. 1999;21:1074–1090. Abstract |
Full-Text PDF (1281 KB)
|
CrossRef
Fenton et al., 1997. 8.Fenton WS, Blyler CR, Heinssen RK. Determinants of medication compliance in schizophrenia: empirical and clinical findings. Schizophrenia Bulletin. 1997;23:637–651. MEDLINE Gray et al., 2002. 9.Gray R, Wykes T, Gournay K. From compliance to concordance: a review of the literature on interventions to enhance compliance with antipsychotic medication. Journal of Psychiatric and Mental Health Nursing. 2002;9:277–284. MEDLINE |
CrossRef
Haywood et al., 1995. 10.Haywood TW, Kravitz HM, Grossman LS, Cavanaugh JL, Davis JM, Lewis DA. Predicting the revolving-door phenomenon among patients with schizophrenic, schizoaffective, and affective-disorders. American Journal of Psychiatry. 1995;152:856–861. Hall et al., 2001. 11.Hall M, Meaden A, Smith J, Jones C. The development and psychometric properties of an observer-rated measure of engagement with mental health services. Journal of Mental Health. 2001;10(4):457–465. Helgason, 1990. 12.Helgason L. 20 years follow-up of 1st psychiatric presentation for schizophrenia — what could have been prevented. Acta Psychiatrica Scandinavica. 1990;81:231–235.
CrossRef
Hogan et al., 1983. 13.Hogan TP, Awad AG, Eastwood R. A self-report scale predictive of drug compliance in schizophrenics — reliability and discriminative validity. Psychological Medicine. 1983;13:177–183. MEDLINE |
CrossRef
Kampman and Lehtinen, 1999. 14.Kampman O, Lehtinen K. Compliance in psychoses. Acta Psychiatrica Scandinavica. 1999;100:167–175.
CrossRef
Kay et al., 1987. 15.Kay SR, Fiszbein A, Opler LA. The positive and negative syndrome scale (PANSS) for schizophrenia. Schizophrenia Bulletin. 1987;13:261–276. MEDLINE Leucht and Heres, 2006. 16.Leucht S, Heres S. Epidemiology, clinical consequences, and psychosocial treatment of nonadherence in schizophrenia. Journal of Clinical Psychiatry. 2006;67:3–8. Mcevoy et al., 2006. 17.Mcevoy JP, Johnson J, Perkins D, Lieberman JA, Hamer RM, Keefe RSE, et al. Insight in first-episode psychosis. Psychological Medicine. 2006;36:1385–1393. MEDLINE |
CrossRef
Morisky et al., 1986. 18.Morisky DE, Green LW, Levine DM. Concurrent and predictive validity of a self-reported measure of medication adherence. Medical Care. 1986;24:67–74. MEDLINE |
CrossRef
Nosé et al., 2003. 19.Nosé M, Barbui C, Tansella M. How often do patients with psychosis fail to adhere to treatment programmes? A systematic review. Psychological Medicine. 2003;33:1149–1160. MEDLINE |
CrossRef
Olfson et al., 2000. 20.Olfson M, Mechanic D, Hansell S, Boyer CA, Walkup J, Weiden PJ. Predicting medication noncompliance after hospital discharge among patients with schizophrenia. Psychiatric Services. 2000;51:216–222. MEDLINE |
CrossRef
Perkins, 1999. 21.Perkins DO. Adherence to antipsychotic medications. Journal of Clinical Psychiatry. 1999;60:25–30. Perkins, 2002. 22.Perkins DO. Predictors of noncompliance in patients with schizophrenia. Journal of Clinical Psychiatry. 2002;63:1121–1128. MEDLINE Piatkowska and Farnhill, 1992. 23.Piatkowska O, Farnhill D. Medication — compliance or alliance? A client-centred approach in increasing adherence. In: Kavanagh DJ editors. Schizophrenia. An overview and practical handbook. London: Chapman & Hall; 1992;. Robinson et al., 1999. 24.Robinson D, Woerner MG, Alvir JMJ, Bilder R, Goldman R, Geisler S, et al. Predictors of relapse following response from a first episode of schizophrenia or schizoaffective disorder. Archives of General Psychiatry. 1999;56:241–247.
CrossRef
Thompson et al., 2000. 25.Thompson K, Kulkarni J, Sergejew AA. Reliability and validity of a new Medication Adherence Rating Scale (MARS) for the psychoses. Schizophrenia Research. 2000;42:241–247. Abstract | Full Text |
Full-Text PDF (82 KB)
|
CrossRef
Young et al., 1986. 26.Young JL, Zonana HV, Shepler L. Medication noncompliance in schizophrenia: codification and update. Bulletin of the American Academy of Psychiatry and the Law. 1986;14:105–122. a Department of Psychology, Institute of Psychiatry, King's College London, University of London, UK b Biostatistics Group, School of Epidemiology & Health Sciences, University of Manchester, UK c Department of Psychiatry and Behavioural Sciences, Royal Free and University College Medical School, University College London, University of London, UK d School of Medicine, Health Policy and Practice, University of East Anglia, UK  Corresponding author.
PII: S0920-9964(07)00500-2 doi:10.1016/j.schres.2007.10.029 © 2007 Elsevier B.V. All rights reserved. | |
|
FULL TEXT