The minimum clinically important difference on the sleep disorders inventory for people with dementia

Sleep disturbances in dementia causes distress to people with dementia and their family carers and are associated with care home admission. The Sleep Disorders Inventory (SDI) is a validated questionnaire of sleep disturbances in dementia often used to measure treatment effectiveness, but the minimum clinically important difference (MCID) is unknown.


| INTRODUCTION
There are currently an estimated forty seven million people worldwide living with dementia, with projections that this number will triple by 2050 due to increasing life expectancy. 1 Sleep disturbances are common in people with dementia, 2 with meta-analyses finding the prevalence in people with Alzheimer's disease was 39%, 3 and in people with dementia living in care homes was 38%. 4,5 When an individual with dementia has disturbed night-time sleep, this can impact on other family members and often means their sleep is also interrupted. 6 This is associated with family carers developing depressive symptoms, and a higher likelihood that they become unable to continue caring at home and the person living with dementia is moves into a care home. 7,8 Currently there are no known efficacious treatments for sleep disturbances for people with dementia. [9][10][11] Emerging evidence suggests multicomponent interventions including a combination of light therapy, cognitive behavioural therapy, and sleep hygiene 2 may prove to be effective, given that the causes of sleep disturbance in dementia are often multifactorial and relate to brain changes from the illness but also to discomfort, pain, anxiety and lack of daytime light and activity. 12 Accurately measuring sleep disturbances in dementia is important for assessing the efficacy of treatments, however there is debate over whether questionnaires or actigraphy are the gold standard of measurement in people with dementia. Actigraphy, where an actigraph is worn on the wrist and measures movement, infers sleep from a lack of movement. 13 It is described as an objective measure, though it is not a direct measure of sleep disturbances and cannot accurately measure daytime sleepiness. 4,13 Furthermore, it is also common for people with dementia to remove their actigraph, accounting for the exclusion of a third of actigraphy data in studies with people with dementia. 13 Questionnaires, on the other hand, are cheaper than actigraphy, allow for more data to be collected, and report on a broader range of sleep disturbances. 14 However, they can be difficult as people with dementia often are unable to remember how they slept and therefore specific proxy rated questionnaires are commonly used. 4 The only validated dementia specific questionnaire measuring sleep disturbance is the Sleep Disorders Inventory (SDI), which can be used as a proxy measure. 15,16 This scale comprises seven different sleep disturbances common in people with dementia, including getting up during the night, getting involved in inappropriate activities during the night, and excessive daytime sleepiness (see Table 1). Each item is rated by the frequency of how often it occurs (1-4) and how severe the disturbance is (1-3), with both scores multiplied together to give a potential item score of between 0-12, with a higher score meaning more frequent and severe sleep disturbance. 15 A score of ≥4 on any item indicates a clinically significant symptom, with the total SDI score (0-84) generated by adding the scores of individual items.
The minimal clinically important difference (MCID) is the smallest change after an intervention that is considered meaningful or valuable by patients and their families. [17][18][19] As the MCID will indicate clinical meaningfulness, it is an additional and possibly more useful measure of the effectiveness of an intervention than statistical difference. 20 To our knowledge no MCID has currently been reported for the SDI or any instrument to measure sleep disturbances in dementia. Therefore, we aimed to derive the MCID of the SDI using three different approaches.

| DREAMS START
This study used data collected from the DREAMS START project (Dementia RElAted Manual for Sleep; STrAtegies for RelaTives) to determine the MCID on the SDI. DREAMS START was a randomised control trial of the feasibility and acceptability of a multicomponent intervention for people with dementia and sleep disturbances. 21

| Determining the MCID
Lassere at al 24 proposed three approaches for classifying the MCID: the distribution, anchor and Delphi approaches.
1. The distribution approach determine the level of change that is required to demonstrate that a change in an outcome measure after intervention is more than would be expected from chance. 17 A SD (SD) of 0.33, 0.4 or 0.5 can be classified as MCID, 20 though it has also been suggested that a value of 0.5 should be used as the default. 25 2. The anchor-based method determines the MCID by comparing the change in the scale of interest and a different scale which measures improvement. 17 In this study we used the DEMQOL-Proxy measure of quality of life as the anchor. 23 It is a validated and frequently used quality of life measure, which is specific for dementia and is completed by a proxy informant.
3. The Delphi method involves the presentation of a questionnaire, in this case about meaningful change in sleep disturbance in a person with dementia, to a panel of individuals with expertise in a field in order to obtain a consensus. 26 There may be several rounds and the process ends when a consensus is reached. 27 Ten to fifteen participants are enough to reach a consensus decision. 28

| Anchoring Results
The SDI change and DEMQOL-Proxy change correlation = −0.01; P = 0.958. Table 1 shows that the only significant correlation between change in individual items on the SDI and change in the DEMQOL-Proxy was an improvement in individuals who had previously got up at night and tried to go out. Table 2 shows the demographic information for the 12 participants who took part in the Delphi questionnaire. The first round did not come to a consensus and participants made suggestions about clarifying the changes described in sleep disturbances so there were clearly different outcomes, which would help them to choose which option qualified as MCID. We then revised the questionnaire was then revised based on these comments.

| Second Round of Delphi
We sent a follow up questionnaire. For other health conditions, the MCID on specific sleep questionnaires have also been defined using similar methods. For example including rheumatoid arthritis using the anchor and Delphi process, 31 in Parkinson's disease using the anchor and distribution methods, 32 and for insomnia using an anchor question that was asked directly to participants about what was a MCID to them. 33 We were unable to do this to define the MCID as this data was not collected in the DREAMS START trial.
Overall in dementia studies, a systematic review found that in trials testing treatments to slow or stop the progress of dementia, only 46% used MCID's for the main outcomes used. 34 Of those 46% of studies, most tended to use an already established MCID for a cognitive scale or used the anchoring method, and none of them used a Delphi asking participants for their opinions of risks vs benefits of drug treatments in dementia. 34 It is important for trials to consider the opinions of people with dementia and their family carers, not just clinicians, 34 as the MCID should give the certainty that a treatment is benefitting the patient. 19 Later studies in dementia such as the Domino trial 20

| Strengths and Limitations
The sample of people with dementia was mixed in terms of diagnosis, age, severity of dementia and ethnicity, and all lived in their own homes therefore results may be generalisable to populations in higher income countries. We used three different methods to explore the MCID results in different ways and we have been transparent in reporting our reports, so that other investigators in different studies can use the results which are most appropriate to their study as is recommended. 29 The distribution method is advantageous because it has the ability to account for change beyond a level of chance. 17  On the other hand, anchor-based methods will produce different MCID values depending on the subjective choice of the anchor. 19 The Delphi consensus process required two rounds of questionnaires. The participants came from a range of different backgrounds and had a varied mix of demographic characteristics but there were only 12 participants, and a larger sample with more people with dementia and family carers would have been beneficial.

| CONCLUSIONS
It is important to deliver cost effective care which improves outcomes for people with dementia in clinical practice, 37 including for outcomes that can affect many aspects of a person with dementia's life, such as sleep disturbance. The results from the present study will help to understand whether interventions aimed at improving sleep disturbances people with dementia are clinically meaningful to the individual, regardless of if results are statistically significant. The results indicate that a decrease of 4 points on the SDI is considered the minimum value to be a meaningful and worthwhile change for the patient, derived from both distribution and Delphi consensus methods. We were unable to use the anchoring method as our anchor, quality of life, was not associated with overall sleep disturbances. As the SDI is currently the only validated dementia specific sleep disturbance questionnaire, and is being used in recently published studies measuring sleep in people with dementia including a randomised controlled trial, 30,38-41 these results will also be helpful for researchers to use in future studies.