Evaluating the Effects of Music Interventions on Hospitalised People With Dementia

RATIONALE: Dementia presents a significant health problem throughout the world. We will compare the effects of a music therapy intervention and an occupational therapy intervention in a randomised controlled trial conducted with hospitalised older people with dementia.

The study will employ physiological, psychometric and observer-rated measures to monitor change over time in a range of parameters including cognitive and behavioural responses, mood and agitation, CNS function, facial expression and blood pressure. The battery of psychometric instruments consists of well-validated and reliable measures. Cognitive status will be measured using The Mini-Mental State Exam (Folstein et al, 1975) and the Hierarchic Dementia Scale (Cole and Dastoor, 1996) Mood will be assessed using the Geriatric Depression Scale (Yesavage et al., 1983). A wide range of behaviours will be assessed with the Cohen-Mansfield Agitation Inventory (CMAI) (Cohen-Mansfield, J., 1986). To monitor changes related to the music and non-music interventions, we will employ the EEG methods that will detect changes in dementia-related generalised, global slowing of the EEG as well as more localised changes at sites in anterior and posterior regions of both cerebral hemispheres. EEG-coherence methods will be used to assess intrahemispheric and interhemispheric functional interactions. We will also record auditory evoked potentials to evaluate changes in the attentional status of participants. This will provide data that will detect change over time in response to the interventions. The Facial Activation Coding System (Ekman \& Friesen, 1978) allows systematic analysis of facial expressions from video-taped recordings. It has been used with people with dementia to provide valid and reliable quantitative estimates of pain states (Hadjistavropoulos et al., 2000). It will be employed here to quantitatively evaluate changes in facial expression that occur during the intervention sessions. This will allow us to observe within group changes and between group differences in the range of expressive responses. Changes in blood pressure will provide a means of evaluating physiological responses to the interventions.

AIMS: The primary aim of the present study is to determine whether the symptomatic improvements attributed to music therapy are supported by empirical evidence derived from a battery of psychometric and observational measures. A second aim is to determine whether neurophysiological measures (EEG frequency, EEG-coherence, EP) and vascular activity (blood pressure) reflect central nervous system and autonomic changes that directly correlate with reductions in dementia-related symptoms.

Specifically, the study will investigate the following hypotheses -

1. Music employed as a therapeutic medium will reduce symptomatic cognitive, behavioural, emotional and functional deficits associated with dementia.
2. Symptomatic changes will correlate with physiological changes. METHODOLOGY: This study will take advantage of an established collaborative network between the Department of Psychiatry, The University of Melbourne, The National Ageing Research Institute, The Cognitive, Dementia and Memory Service (CDAMS) at The Royal Melbourne Hospital and the CDAMS at the St George's Hospital campus of St Vincent's Health. The research team covers several disciplines including aged psychiatry, neuropsychiatry, geriatric medicine, neuropsychology, behavioural neuroscience, electrophysiology, music and music therapy.

SUBJECTS: One hundred and eighty sub-acute patients with mild, moderate and severe dementia will be recruited from RMH (Royal Park Campus) and St Georges Hospital over a period of 18 months. There will be a potential pool of approximately 800 patients over the planned 18 month data collection period. We anticipate that with ineligibility of 30%, an attrition rate of up to 20-25% and time lost to holidays and a participation rate of 70%, our total cohort will be approximately 260, leaving substantial redundancy over the numbers required to power the study.

Testing will commence when each patient has settled for one week when the majority of illnesses are stabilised and medications adjusted. A full medical assessment will be available for all participants in order to evaluate medical comorbidity, cognitive status and dementia diagnosis. This information will include results of physical examination, blood tests, Ct scan, informant history and neuropsychological assessment which includes the administration of the MMSE (Folstein et al. 1975) and classification according to DSM-IV, ICD-10 and NINCDS-ARDRA criteria, the current research gold standard for dementia diagnosis (American Psychiatric Association, 1994). Consensus diagnosis with information available will be made by the clinicians involved in the study. Primary diagnoses, comorbidities and medications will be monitored. Recruitment will be undertaken in accordance with procedures approved by The Mental Health Research and Ethics Committee, Melbourne Health Network, the St Vincent's Health Research Ethics Committee. Given the nature of our participants' diagnoses, it is recognised that they cannot give informed consent.

PROCEDURE: Participants will be randomly assigned to either the music intervention group or the control (occupational therapy) group. Both groups will undertake three 50 minute intervention sessions per week for three weeks giving a total of nine interventions. Each session will be conducted with groups of 4 - 6 participants.

Pre and post measures will be made on the day prior to the therapy session and the day after the last therapy session. These measures are as follows:

• EEG recording made that will yield slow/fast ratio data and resting state and music-listening EEG-coherence data and auditory evoked potential data.
• Four blood pressure and pulse readings taken at hourly intervals to establish a mean baseline.
• Questionnaires - Mini Mental State Examination, Hierarchic Dementia Scale, Geriatric Depression Scale will be administered prior to the first intervention session and after the ninth intervention session.

Pre and post testing involving will take approximately 90 minutes in total. Video recordings will be made using two cameras during the fifth therapy session. The first camera will be located on a tripod such that the whole group, including the therapist, will be continuously in frame to facilitate subsequent observations about participation and social interactions. The second camera will be hand held and will take film of the faces of each participant during the session for subsequent coding of changes in facial expressions using the Facial Activation Coding System.

Further information will be compiled by members of the research team. This information, used to monitor comorbid conditions and pre and post changes, will be derived from patient records. It will be collected the following instruments

• The Clinical Dementia rating Scale (Hughes et al, 1982)
• The Barthel Index (Mahoney and Barthel, 1965)
• The Cumulative Illness Rating Scale (Linn et al 1968).
• The Neuropsychiatric Inventory (Cummings, 1997) Finally, The Cohen-Mansfield Agitation Inventory (Weiner et al, 1998) will administered pre and post intervention by a researcher in consultation with nursing staff.

In addition to the above measures, the full range of treatments outside the music therapy interventions received by participants in both groups will be documented. This will enable the researchers to monitor the levels of therapist contact experienced by participants in both the music therapy and control groups. It will also provide data that may be used as covariates in a number of statistical analyses.

INCLUSION CRITERIA:People with a clinical diagnosis of dementia who are currently in sub-acute aged care hospital wards; EXCLUSION CRITERIA: delirium; acute medical conditions (eg pneumonia and cardiac failure) where it is likely that participation is contra-indicated; primary psychiatric conditions; severe hearing impairment; RANDOMISATION PROCEDURES: A stratified randomization procedure will be used to ensure that both the experimental and control groups are at least broadly homogeneous with respect to cognitive status. The procedure will be applied separately within each of the two hospital facilities with allocation made progressively as people enter the study. After consent has been acquired and the pre-intervention measures taken, participants will be coded, stratified into two levels according to MMSE scores then allocated to treatment groups according to the computer generated random sequences. The sequences will be generated using research randomiser, an online web-based program (http://www.randomizer.org/form.htm). The person responsible for implementing the allocation sequence will be either a member of the hospital staff or a NARI staff member who has no involvement in any other aspect of the study. That person will be responsible for de-identifying participants using a conventional coding system. Baseline and post-intervention data collection procedures will be administered by persons who have no other role in the project procedures and will remain blind to treatment allocation. Baseline testing will take place before allocation. As they will have no contact with participants during the intervention procedures and because patient identification is coded by an independent person, the blind can be maintained. As added protection, psychometric instruments (MMSE, HDS and GDS) will be assessed only once by an assessor for each participant. That is, if Assessor A administers the instruments with Participant A for baseline testing, Assessor B will administer the instruments post-intervention for Participant A. This will avoid any unintended prior knowledge of individual participant performance and so help to ensure post-treatment blinding with respect to active versus control group assignment. The instruments to be used have clear and established implementation procedures with high inter-rater reliability. After the post-intervention assessments have been taken for each participant, the assessors will be asked to nominate which group they thought the participant was in. A simple statistical analysis will be able to detect whether there is a significant bias in group assignment or whether assignment is no better than chance. Bias may arise, for example, if one treatment group has an outcome that strongly differentiates it from the other.. It is noted that, while there are other potential stratification factors (eg age, gender, diagnoses), the use of more factors rapidly becomes unfeasible. Because current cognitive status has some prognostic relevance, particularly with regard to music therapy, it has been decided to use that measure as the sole stratification factor.

STATISTICAL ANALYSES: Given the multiple dependent variables being monitored in this study, a MANOVA will be used to test for between-group differences with post-hoc univariate comparisons as required. Based on previous studies at NARI referred to above using similar measures we expect that 90 subjects per group will detect a 25% change in mean values with at least 80% power (alpha .05). For test-retest data, a two way repeated measures MANOVA (group, time) will be used to examine changes over the duration of the intervention, group effects and the interaction between these factors. Bivariate correlations and multiple regression analyses will be used to explore the relationships between the various measures (EEG, FACS, MMSE, HDS, self-report and observational measures). Similar comparisons can also be made for change scores within each domain of parameters (objective, psychometric, self-report) and the way they co-vary over time.