In Japan, where older people already make up more than 23% of the population and the proportion is still growing, the burden on those caring for people with dementia is an increasing problem. This burden is magnified by wandering behavior, a peripheral symptom. Thus, there is a need for an objective measure of wandering behavior to determine what constitutes effective care. In this study, we translated the Algase Wandering Scale – Version 2 into Japanese, and examined its reliability and validity. Ambulatory residents with dementia were selected from two nursing homes and two wards specializing in dementia care in hospitals in Japan. Nurses and care workers taking care of these residents answered questionnaires regarding the residents. From the results, the Algase Wandering Scale – Version 2, Japanese version, was examined for inter-rater reliability, stability, internal consistency, and concurrent validity. The results of the analysis in the present study demonstrated that the Algase Wandering Scale – Version 2, Japanese version, has reliability and validity, and that it can measure the presence or absence of wandering and its severity. Surveys of residents with various wandering patterns in many facilities and verification of construct validity are warranted in the future.
With older people accounting for 23.3% of the population, aging has advanced more in Japan than in any other country making it a super-aged society. Moreover, the proportion of the population that is elderly is expected to continue rising in Japan, reaching 39.9% in 2060 (Japanese Cabinet Office, 2012). The incidence of dementia in older people in Japan is said to be 8.3% (Sekita et al., 2010), and as aging continues, further increases are predicted in the number of people diagnosed with dementia.
In caring for older people with dementia, peripheral symptoms of dementia are a strong contributor to increasing the care burden (Greiner et al., 2007). Wandering behavior is one such burden, as it means that caregivers must maintain a constant watch (Chiu et al., 2011). This puts significant time constraints on the caregiver and causes a risk of fractures for the older person due to falls, which in turn might lead to the person becoming bedridden (Algase et al., 2004; Son et al., 2006). There is a need for investigations of measures to deal with wandering. To understand wandering behavior, evaluating related factors and care is important for both the physical and mental stability and safety of older people and an objective measure is required.
Reports on the incidence of wandering range from 17.4% (Klein et al., 1999) to 36% (Logsdon et al., 1998) in the community, and from 11% to 24% (Lucero et al., 1993) or 39% (Cohen-Mansfield et al., 1991) in institutions. Wandering is understood in various ways by different researchers: from an adaptive behavior that provides physical stimulation and exercise (Cohen-Mansfield et al., 1997) to a stress-coping behavior (Matteson & Linton, 1996), self-stimulation behavior (Lucero et al., 2001), and a behavior to which difficulty in adapting to the environment is a contributing factor (Futrell & Melillo, 2002). Wandering is generally considered to be a behavior that has no aim (Dawson & Reid, 1987; Hussian & Brown, 1987; Hope & Fairburn, 1990; Madson, 1991; Matteson & Linton, 1996; Holmberg, 1997; Heard & Watson, 1999; Kiely et al., 2000; Colombo et al., 2001; Rolland et al., 2005), however there are also people who view it as a goal-directed behavior, a behavior that fulfills some kind of need (Thomas, 1995; Beattie & Algase, 2002), and a behavior in which having and not having a goal coexist (Snyder et al., 1978; Futrell & Melillo, 2002). Algase et al. (2001a), who developed the Algase Wandering Scale (AWS), analyzed past studies and defined wandering behavior as a walking behavior consisting of five dimensions: frequency, pattern, boundary transgressions, deficits in navigation or wayfinding, and temporal aspects. Algase et al. (1996; Algase, 1999) view behavioral and psychological symptoms of dementia (BPSD), including wandering behavior, as a behavior that is produced by a latent need, and they proposed the need-driven dementia-compromised behavior model, in which wandering was taken to be an interaction between background and proximal factors.
Based on these definitions and models, Algase developed the AWS to measure wandering behavior (Algase et al., 2001a). Surveys of wandering behavior by direct observation (Snyder et al., 1978; Cohen-Mansfield et al., 1991; Goldsmith et al., 1995; Matteson & Linton, 1996; Algase et al., 1997; 2001b) or video recording (Martino-Saltzman et al., 1991; Lucero et al., 1993; Edgerly & Donovick, 1998) require much time and labor, and the AWS was developed as a way to overcome this (Algase et al., 2001a). The AWS has five subscales (persistent walking [PW], spatial disorientation [SD], elopement behavior [EB], routinized walking [RW], and shadowing [SH]) consisting of 28 items. Cronbach's alpha (internal consistency) of the AWS was 0.86, and the inter-rater reliability was 0.65 (P < 0.01). However, sufficient reliability was not obtained for the subscales of RW and SH. Validity was examined using concurrent validity and factor analysis. The five factors described above were extracted, and factor loadings of all items were greater than 0.4. Bartlett's test of the sphericity of AWS was significant (P < 0.001), and the Kaiser–Meyer–Olkin measure of sampling adequacy was 0.84.
A revised version, the AWS – Version 2 (AWS-V2), was then developed (Algase et al., 2004). The AWS-V2 was originally a four-step scale with 38 items. A factor analysis of these 38 items was conducted, and, finally, five factors and 27 items were adopted. The reliability and validity of these 27 items were investigated, and it was found that, while the subscale of SH was strengthened compared with the AWS, RW was not improved. Moreover, sufficient reliability and validity were not reached for attention shifting, which was a new addition. In our study, a new AWS-V2, with further revisions in which attention shifting was returned to RW, was obtained directly from Dr Algase, and a Japanese version was prepared.
The purpose of this study was to translate the AWS-V2 into Japanese and examine its reliability and validity.
This study used a cross-sectional, correlational design. For the examination of reliability and validity, two surveys were conducted. The Japanese version of the AWS-V2 was examined for inter-rater reliability and test–retest reliability in survey 1, and inter-rater reliability, test–retest reliability, and concurrent validity in survey 2. Survey 1 was conducted focusing on the verification for reliability of raters' multiplicity, and survey 2 was implemented to confirm validity and consider the multiplicity of wandering pattern and level.
Settings and sample
Ambulatory residents with dementia were recruited from two nursing homes and two wards specializing in dementia care in hospitals in the Chubu and Kanto regions, Japan. Nurses and care workers taking care of these residents answered questionnaires regarding the residents. Because residents in these facilities were cognitively impaired, consent was obtained from proxies authorized to make care and medical decisions. Residents with dementia who were regarded as ambulatory by the manager of their respective facility or hospital and for whom proxy consent could be obtained were included.
We obtained approval from the Research Ethics Committee at the Graduate School of Medicine, Osaka University for survey 1, and the School of Nursing, Mie Prefectural College of Nursing for survey 2, both institutions in Japan. Explanations of the request were given to the proxies of the residents; these explanations included the purpose and method of the study, privacy protection, and a statement that participation was voluntary with no penalty for non-participation.
Translation of the AWS-V2
Permission for translation of the AWS-V2 was received directly from Dr Algase, the scale developer. During the English-to-Japanese translation process, the first author consulted Dr Algase about ambiguous phrases. After back-translation from Japanese to English, the meanings of each item of the original and Japanese versions were compared by the scale developer. Modification and comparison were continued until consent for all items was obtained from the developer.
The AWS-V2 was developed by Algase et al. (2004) to measure the wandering behavior of residents with dementia living in facilities. The AWS-V2 includes 23 items as a wandering scale, and five other items for verifying reliability and validity. The scale is constructed from five subscales: PW, SD, EB, SH, and RW. The scale uses a four-point ordinal scale for each item, and the total score ranges from 23 to 92. A higher score indicates more frequent wandering behavior.
Mini Mental State Examination
The Mini Mental State Examination (MMSE) is used to screen for cognitive function impairment. The scores range from 0 to 30; a score below 24 indicates cognitive impairment (Folstein et al., 1975). This examination produces a global performance score from 11 items that measure orientation, registration, attention and calculation, recall, language, and construction tasks.
The Gottfries–Bråne–Steen (GBS) Scale is a dementia symptom-evaluation scale constructed from four subscales: motor, intellectual, emotional functions, and symptoms characteristic of dementia syndromes (Gottfries et al., 1982). Each item is rated on a seven-point scale from 0 to 6, and the score ranges from 0 to 156. A higher score indicates worse symptoms.
Behavioral Pathology Alzheimer's Disease Rating Scale
Behavioral Pathology Alzheimer's Disease (BEHAVE-AD) Rating Scale was developed as a scale to evaluate the effects of Alzheimer's disease medications (Reisberg et al., 1987; 1996). The BEHAVE-AD Rating Scale is based on 26 items concerning eight factors: paranoid and delusional ideation, hallucinations, activity disturbances, aggressiveness, diurnal rhythm disturbances, affective disturbance, anxieties and phobias, and a global rating. Each item is rated on a four-point scale from 0 to 3, and the score ranges from 0 to 78. A higher score indicates a worse condition. The BEHAVE-AD Rating Scale was designed to be particularly useful in prospective studies of behavioral symptoms in patients with Alzheimer's disease (Burns et al., 2004).
Questionnaires were distributed to staff taking care of residents with dementia. The completed questionnaires were put into envelopes, sealed, and posted to a collection box placed in the wards.
Age, sex, occupation, and length of work experience were collected as staff demographic data. Cognitive impairment of the residents with dementia was assessed using the MMSE.
Data analysis: survey 1
All staff members were required to evaluate 12 residents with dementia using the AWS-V2, and interclass correlation coefficients were examined using these respondents' data.
One month later, a second AWS-V2 measurement was conducted on 12 residents by the staff members who performed measurements and responded to the initial questionnaire. Test–retest reliability was estimated using intraclass correlation coefficients.
Data analysis: survey 2
For each of the 40 residents with dementia, two staff members conducted AWS-V2 evaluations of that resident, and inter-rater reliability was estimated using interclass correlation coefficients.
One month later, a second AWS-V2 measurement of 56 residents was conducted by the same staff members who conducted the first measurement. Test–retest reliability was estimated using intraclass correlation coefficients.
Internal consistency was measured using Cronbach's alpha coefficient for the AWS-V2 total and its subscales.
Concurrent validity was examined by measuring the correlation of the AWS-V2 and its subscales with the MMSE, the BEHAVE-AD Rating Scale, the GBS Scale, and scores (corresponding to “Definitely not”; “At times”; “Yes, but it is not a problem”; and “Yes, and it is a problem”) for the single-item indicator “This resident is a wanderer” of AWS-V2. The validity of the AWS-V2 was also estimated by cross-comparing the AWS-V2 and its subscale scores with the differences by level (mild, moderate, or severe) of cognitive impairment based on MMSE scores and the score for the single-item indicator “This resident is a wanderer” using one-way ANOVA.
The interclass correlation coefficients in this scale looked at the intermeasurer correlation between 18 staff members for each resident. These correlation coefficients were calculated with more people than in the past, so the level of acceptability is thought to be sufficient to judge the scale as reliable.
In survey 2, the correlation coefficient was 0.797 in samples 1 and 2. The correlation coefficient between the two samples for the AWS was 0.65 (Algase et al., 2001a), and the inter-rater correlation coefficient of this scale showed a value higher than this. In addition, Algase et al. (2004) showed a kappa coefficient of ≥ 0.40 for 17 of 27 items, and < 0.40 for the remaining items on the AWS-V2. Therefore, an inter-rater reliability was achieved with the present scale equal to or greater than that of the original scale developed by Algase et al. (2001a, 2004).
The test–retest reliability of this scale was 0.889 in survey 1 and 0.863 in survey 2. Reliability coefficients of approximately 0.70 are generally thought to be appropriate (Polit & Beck, 2004), and the fact that this scale exceeded that level suggests that it has adequate reliability. For the subscales, however, SH did not exceed 0.70 in either survey 1 or 2, and issues in stability remain with the retest method for the subscale.
Intra-rater reliability with this retest method was not measured in the process of developing the AWS-V2 (Algase et al., 2004). The stability of the AWS-V2 was examined with this retest method for the first time in this study. The findings suggest that, if the values of the entire scale are used, this scale can be used, even in longitudinal studies, such as the measurement of wandering before and after nursing intervention and during follow-up periods.
This scale showed high consistency, except for SH. The AWS-V2 total and all subscales, except for SH, showed consistency equal to or above the original AWS-V2. In particular, high reliability of six points (over original estimate) was obtained for the AWS-V2 total.
The analysis results for reliability demonstrated high reliability for the PW, SD, EB, and RW subscales.
With the original AWS-V2, the correlation with the MMSE score was investigated, and significant correlations were found with the AWS-V2 total and the PW, SD, and EB subscales (Algase et al., 2004). In our study, significant correlations were seen with the AWS-V2 total and all subscales, and higher validity than with the original version was obtained.
Correlations with the BEHAVE-AD Rating Scale and GBS Scale were also investigated. Significant results for the AWS-V2 total and all its subscales were obtained with both scales, and the validity was verified. The highest correlations were seen with activity disturbances among the subscales of the BEHAVE-AD Rating Scale and with symptoms characteristic for dementia syndromes among the subscales of the GBS Scale. This is thought to be because activity disturbances is the subscale in the BEHAVE-AD Rating Scale for behaviors with a particularly close relationship to wandering, and symptoms characteristic for dementia syndromes is a subscale that measures BPSD in the GBS Scale. These are the subscales with the closest relationship to wandering behavior among the subscales of each scale, and this result also confirms that the Japanese version of the AWS-V2 has high validity.
The results of the correlations and the comparisons for AWS-V2 and wandering level demonstrated the validity of the AWS-V2 total and all subscales, and the results of the correlations and the comparisons for the AWS-V2 and cognitive level demonstrated the validity of subscales other than the AWS-V2 total and the SH and RW subscales. For cognitive level, the result was the same as the original AWS-V2, but for wandering level, a higher validity than the original was obtained in both correlations and post-hoc analyses. Moreover, although PW showed the highest discrimination for whether or not a person is a wanderer in the original scale, in the present scale, SD showed the highest discrimination. In both the original and Japanese versions, SH and RW were unrelated to the level of cognitive function. This suggests, as also stated by Algase et al., that these two subscales emerge without relationship to the level of cognitive function, or perhaps they have weak sensitivity to cognitive function (Algase et al. 2004).
There are several limitations with this study. A factor analysis for the AWS-V2 Japanese version was not performed; therefore, the construct validity was not verified. For the rest–retest reliability, residents were re-examined 1 month after the first survey. During this period, residents might have had behavioral changes, and these changes might have affected the results. In addition, the staff's knowledge and experience in dementia care might have impacted their rating results.
The results of the analysis in this study demonstrated that the Japanese version of the AWS-V2 has reliability and validity, and that it can measure the presence or absence of wandering and its severity. In addition, the stability of the AWS-V2 was verified, and thereby the AWS-V2 was demonstrated to be suitable even for longitudinal studies; such demonstration had not been performed for the original English version.
As this scale focuses more on pattern classification than on the frequency of wandering, future surveys of residents with various wandering patterns in many facilities and verification of construct validity from factor analysis are warranted.
We would like to sincerely thank all the participants at the nursing homes and wards specializing in dementia care in hospitals for their contribution.
Study Design: CG, KM.
Data Collection and Analysis: CG, KM, MS, MY.
Manuscript Writing: CG.