Recruitment and subjects

In Norway, the municipalities are legally responsible for providing domiciliary and residential care, and the administration of nursing homes and day care centres is organized within the municipalities’ public health services. Most patients receive domiciliary care for as long as possible prior to admittance to a nursing home. The municipalities’ health care and other care services in cooperation with patients’ general practitioners assess patients’ need for residential care. Most nursing homes have both ordinary units and special care units, and often include day care facilities. Special care units are adapted units that usually only house 7–8 PWDs.

For our study, the county development centres for dementia care in three counties in the south-eastern part of Norway were responsible in recruiting nursing homes and day care centres in their municipalities. All nursing homes and day care centres in the three counties were invited to participate in the study, and 15 nursing homes with adapted units for PWDs and 23 adapted day care centres for home-dwelling PWDs were willing to participate.

Each participating institution was asked to recruit between 5 and 8 participants. The inclusion criteria were: aged 65 years or older and either a diagnosis of dementia or a cognitive deficit measured as a score less than 25 on the Mini-Mental State Examination test [41–43].

A total of 209 participants were recruited (88 PWDs living in a nursing home (PWD NH), and 121 home-dwelling PWDs). Due to death or because of withdrawal from the study, 16 participants were excluded from the analyses, which meant that data relating to 193 participants (78 PWD NH and 115 home-dwelling PWDs) were included in the analyses. Home-dwelling participants all took part in a day care centre programme at least once per week. The baseline data collection was carried out in winter–spring 2013 (N = 43 (PWD NH = 17, home-dwelling PWDs = 26)), autumn–winter 2013 (N = 78 (PWD NH = 31, home-dwelling PWDs = 47)), and spring–summer 2014 (N = 72 (PWD NH = 30, home-dwelling PWDs = 42)), mainly due to practical limitations and to avoid seasonal biases. In addition, a subsample consisting of 11 PWDs in nursing homes and 8 home-dwelling PWDs that were randomized to control groups (treatment as usual) was assessed for QoL 6 months after baseline.

Assessments and procedures for data collection

The participating patients’ primary nurses scored all psychometric assessments and collected information on the participants’ age, gender, educational level, use of walking aids, and social encounters. Before the project, they all participated in mandatory education on the use of the instruments.

Quality of life (QoL) was measured using the Norwegian version of the Quality of Life in Late-stage Dementia (QUALID) scale [44, 45]. The proxy rating scale consists of 11 items that are rated on a five-point scale. The items are rated by frequency of occurrence, comprising both positive and negative dimensions of concrete and observable mood and performance. Scores are summed to range from 11 to 55, Cronbach’s alpha = .815 (nursing home = .764, home-dwelling = .719). A lower score indicates a higher quality of life.

The Clinical Dementia Rating (CDR) scale, is a 5-point scale used to assess six domains of cognitive and functional performance relating to dementia [46–48]. CDR staging is a valid substitute for a dementia assessment among nursing home residents when rating dementia and determine the severity of dementia [47, 48]. A CDR global score of 0 implies no cognitive impairment, 0.5 = very mild dementia, 1 = mild dementia, 2 = moderate dementia, and 3 = severe dementia. Before the analyses, the CDR scores were recoded into three groups: mild (0, 0.5 and 1), moderate (2), and severe dementia (3).

Actigraphy (ActiSleep+, ActiGraph, Pensacola, US) was used to measure sleep patterns, physical activity levels, and light exposure. ActiSleep + is a validated 3-axis accelerometer, which has approximately the shape and size of a wrist watch and delivers advanced data about the wearer’s movements over time and their exposure to light. The use of actigraphy for monitoring sleep is validated [49], also for dementia patients [50]. The ActiSleep + was worn on the left wrist continuously for 7 days (epoch-length 1 min). Participants were free to remove the ActiSleep + device but were encourage not to do so. Relatives and caregivers were instructed to encourage the participants to put it back on if it had been removed. Before the measurements started, ActiSleep + was introduced orally, visually, and in written form to the participants by their primary nurse, as well as by their relatives and caregivers.

The actigraphy data were processed using the Scoring and Sleep functions of ActiLife, software Version 6.11.2 (ActiGraph, Pensacola, USA), after applying the Wear Time Validation tool. Days with more than 8 h recorded were included in the further analyses in order to ensure that the activity pattern for those days reflected the participant’s typical behaviour pattern. All subjects included in the analysis had at least three valid days and nights.

Total sleep time (TST) is the amount of actual sleep during the night-time, measured in hours. The term ‘wake after sleep onset’ (WASO) defines the amount of time spent awake after sleep has been initiated and before final awakening; it sums all wake epochs in minutes. The default algorithm of ActiLife may have problems with analysing the sleep–wake schedule. For that reason, we manually inspected all awakenings and created a new variable called ‘Number of awakenings > 5 min’. By using a minimum awake time of 5 min, we ensured that the number of awakenings were accurate. ‘Sleep efficiency’ was defined as the number of sleep minutes divided by the total number of minutes when the participant was in bed, and was expressed as a percentage. Because of the challenge of identifying a precise bedtime and getup-up time among the home-dwelling population, a default time-in-bed period was arbitrarily set as 23:00 to 06:00 h. Therefore, in our study, sleep efficiency referred to the minutes of sleep within the default time period, and not the patients’ actual time spent in bed, and below this is referred to as the ‘Sleep during night period’.

Physical activity levels were calculated using the Freedson Adult Cut Points [51] in ActiLife software, and applying a time filter between 08:00 and 20:00 for each monitored day. ActiLife calculates three activity levels based on the frequency and intensity of the movement. These constitute the measure ‘counts’, which are specified as ‘counts per minute’ (cpm). ‘Sedentary activity level’ is time in percentage with no physical activity (standardized cut point value: 0–99 cpm). ‘Light activity level’ is defined as light intensity activity (standardized cut point value: 100–1951 cpm). Activities in this category could, for example, be standing or household activities. ‘Moderate activity’ (standardized cut point value: 1952–5724 cpm) equates to physical activity, such as walking at 4 km/h. The Freedson Adult Cut Points can also include measures of ‘Vigorous’ activity and ‘Very vigorous’ activity, but these were not used in the study because none of the participants scored any activity at this level. The absolute time (minutes) spent on the different activity levels was subsequently expressed as a percentage of the overall monitoring time.

Light exposure was recorded every second and measured in counts, giving ‘lux average counts’, which indicated the participants’ level of exposure to light.

Records of patients’ use of psychotropic medication (antipsychotics, antidepressants, anxiolytics, and hypnotics/sedatives) were collected (from no/yes responses) and a score for number of different types of psychotropic medications (0–4) was constructed.

Ethics

All participants were informed that they could withdraw from the study at any stage. The ActiGraph device worn by the participants did not register what type of activity they engaged in or their localization, and therefore the usage was not considered invasive.

Analyses

All statistical analyses were performed using IBM SPSS Statistics Version 23.0 for Windows (Armonk, NY: IBM Corp). Cronbach’s alpha was calculated for all sum scores. Group differences between PWD NH and home-dwelling PWDs were tested with one-way ANOVA for continuous variables’ and with chi-square for categorical data. Stratified analyses of the three categories of cognitive level derived from the CDR test were conducted for variables showing significant differences between groups.

A multiple regression analysis was used to test the association between institutionalization and QoL. This model was only tested for PWDs with moderate dementia, due to the low number of persons with mild dementia in nursing homes and low number of persons with severe dementia living at home. Age, gender and variables that were significantly different between the two groups of PWDs in the bivariate analysis, namely social encounters, use of walking aids, physical activity level (moderate), light exposure, and medication, were entered into the analysis in order to control for these factors. Before the analysis, dichotomous variables for walking aids (no/yes) and social contacts (< once per week/≥ once per week) were constructed. Collinearity statistics showed acceptable values (max VIF < 2.3). Heteroscedasticity was not observed.

The association between residency and change in QoL was investigated by linear regression analysis of the subsample with follow-up data (n = 19). Change in QUALID was used as the dependent variable, and residency was entered as predictor variable. In order to control for different baseline levels in QoL, the QUALID baseline score was included in the analysis.

Group characteristics

QoL, sleep patterns, physical activity, light exposure, and medication

In the whole sample, PWDs living in nursing homes showed a significantly lower QoL than home-dwelling PWDs (Table 1). They also scored significantly lower on all sleep parameters, except for ‘Wake > 5 min’. PWDs living in nursing homes experienced almost four times less light exposure compared with home-dwelling PWDs. The actigraphy results showed that PWDs living in nursing homes were less active than and showed significantly more sedentary and less active behaviour than home-dwelling PWDs (Table 1). There were also a significant difference in use of psychotropic medication between PWDs living in nursing homes and home-dwelling PWDs, both in the prevalence and number of medications used (Table 1).

Regression analysis

When looking at change in QoL over time, residency explained 25 % of the change in QoL among participants with moderate dementia (F(2,16) = 3.993, p = 0.039, R² Adjusted = .25). Nursing home PWDs’ mean change in QoL was 1.73, while home-dwelling PWDs’ mean change was −0.38. The analysis shows that both baseline score on QUALID and institutionalization did significantly predict change in QUALID during 6 months (Beta = −0.63, t(19) = −2.66, p < .05) and (Beta = −0.51, t(19) = −2.16, p < .05) (data not shown).

Strengths and limitations

This study had several limitations. The institutions and the participants were all recruited to participate in two RCTs and might not have been representative of nursing home and home-dwelling populations in general. However, the recruitment procedure and inclusion criteria were the same for both patient groups, which made the groups comparable. Furthermore, the participants’ characteristics and level of QoL were in line with findings in other Norwegian studies of persons with dementia [28, 57, 65], indicating that the sample was representative. All home-dwelling PWDs had activities at a day care centre at least once per week, which means that the sample was not representative of the home-dwelling PWD population as a whole. It should be noted that activities at Norwegian day care centres are usually offered to persons who exhibit higher levels or more severe symptoms of dementia, to ease the burden on the family carers.

Validated psychometric questionnaires were used to measure cognitive impairment (CDR) and QoL (QUALID), but proxy assessments always will have less validity than self-assessments. Assessing QoL in PWDs is often done with proxy assessments due to the assumption that the respondents will not able to complete a self-report. However, research has shown that especially those with mild or moderate dementia are capable of completing such reports, and interestingly PWDs in general report that they have a better QoL than their close relatives or care workers do [21, 26]. Therefore, we cannot know whether the scoring in our study truly reflected the participants’ experience, but the QUALID assessment rates concrete and observable behaviour and is a validated tool for this patient group [45]. In our study the patients’ primary nurses scored the psychometric measures, and primary nurses at nursing homes could be expected to have a broad understanding of their patients’ cognitive level, behaviour, and mood. However, the nurses at the day care centre only saw their patients once or twice each week and might not have had the same insights into their patients’ lives. This could have led to bias if the scoring relating to QoL in home-dwelling PWDs was systematically higher than for nursing home PWDs. However, a previous study found no clinically relevant differences in proxy-reported QoL between those in home care and those in institutional long-term care [21].

Actigraphy gives objective data on physical activity. The use of wrist ActiGraphs has been validated for the evaluation of sleep in patients with dementia [66], and the devices have been used to study motor activity patterns in elderly patients with dementia both living in nursing homes and living at home [67, 68]. A specific problem of using actigraphy in this particular patient group could be the removal of the wrist device. We excluded days that had less than 8 h recorded after applying ActiLife’s Wear Time Validation tool (21.75 % of the total number of days); hence, removal of the wrist ActiGraph was not considered a substantial problem. Also, we had to rely on an arbitrary setting of ‘time in bed’, which on the one hand allowed us to obtain a more standardized measure of the amount of sleep during more ‘desirable’ hours of the day or night, but on the other hand did not fully return a measurement of the participants’ actual sleep efficacy. This would have to be taken into account if the findings were to be compared with those in the existing literature.

The model explained 28 % of the variance in QoL, which indicates that other factors not included in the study affected QoL in the patient group. It is reasonable to assume that the PWDs living in nursing homes had poorer health in general, and more comorbid somatic diseases, as might be indicated by the significant differences between the groups in medication and use of walking aids. However, comorbidity and number of diagnoses were not found associated with QoL in another Norwegian study of patients with dementia in nursing homes [31], and earlier research has shown inconsistent associations between medication and QoL in PWDs [21, 31, 40].

Finally, the cross-sectional design of the study restricted the possibility of drawing any conclusions on causality, and the only conclusion that could be derived from the regression model is that living in nursing home is associated with lower QoL. However, the results of the longitudinal analysis conducted on the subgroup population could suggest that residential care contributes in a causal way toward lower QoL.