Factors Associated With Wandering Behaviors in Veterans With Mild Dementia: A Prospective Longitudinal Community-Based Study

Abstract

Objective:

To investigate baseline factors associated with caregiver-reported wandering among community-dwelling veterans with mild dementia.

Methods:

Veterans with mild dementia (N = 143) and their caregivers participated in a 2-year prospective longitudinal study. Measures assessed wandering, daily function, behavior, cognition, and personality features. Wandering was dichotomized as present or absent across study periods, and associations with baseline characteristics were examined.

Results:

One-quarter of participants demonstrated caregiver-reported wandering at 1 or more study visits, with 14% to 15% wandering at any 1 visit. Wandering was associated with significantly lower baseline scores in performance of daily function, behavioral response to stress, gait, and balance, and conscientiousness.

Conclusions:

This novel study evaluated wandering in a community-dwelling sample of veterans with mild dementia. Wandering was associated with a specific personality trait, poorer behavioral response to stress as well as greater functional and gait/balance impairment. These findings may assist in developing community-based interventions for caregivers.

Current Study

The Veterans Health Administration provides care and support to a large and growing number of older veterans with dementia. Dementia care is costly, demanding, and fraught with the potential of negative consequences, particularly among individuals exhibiting wandering behaviors. “Wandering” has been defined as a syndrome of frequent, repetitive, seemingly aimless, typically temporally, and/or spatially-disoriented ambulation in persons with dementia. ^1,2 Efforts directed at the early identification of and intervention for wandering could prevent or at least delay the high cost of care and improve the quality of life for persons with dementia and their caregivers.

This study involved longitudinal assessments of physical health, functional abilities, and behavior among a sample of veterans with dementia living in their homes in the community. Factors shown to be associated with the course of neuropsychiatric symptoms ³ (eg, personality characteristics, behavioral symptoms, and daily functioning abilities), were considered as possible risk factors for wandering. This study (1) collected information at each visit as to whether the caregiver would describe the participant as having wandering behavior and (2) examined whether participant characteristics assessed at baseline were related to caregiver-reported wandering across the 2-year period of observation.

The aims of this study were to (1) document prevalence of caregiver-reported wandering in a cohort of community-dwelling veterans with early dementia and (2) examine baseline demographic, personality, functional, and clinical factors and their associations with caregiver-reported wandering.

Introduction

Wandering behavior is recognized as one of the most challenging issues for caregivers of persons with dementia in both institutional and home settings. It may be a significant source of caregiver distress and lead to institutionalization ⁴ as well as increase the likelihood of negative consequences such as falls and accidents. ⁵ Most studies of wandering have involved individuals with moderate to severe dementia residing in long-term care settings. Few studies of wandering are focused on those with mild dementia who are community-dwelling, yet the community setting may offer an important window of opportunity to intervene and prevent future institutionalization among otherwise healthy persons with mild dementia who have wandering behaviors. Unfortunately, a systematic review of behavioral intervention trials to reduce wandering in dementia failed to reveal evidence to recommend a specific treatment intervention for this challenging set of behaviors. ⁶

The community setting may have unique characteristics such that preventive interventions may be successful. For example, a recent meta-analysis by Brodaty and Arasaratnam ⁷ examined interventions provided specifically by family members for dementia-related behaviors. The analysis suggested the potential of greater success when family caregivers direct the interventions, as would most often occur in community-based trials. The study reported here sought to investigate the characteristics of factors associated with wandering in community-dwelling persons with mild dementia who received family caregiver support. Specifically, this study examined whether the individual was perceived as a wanderer and how this perception of the patient related to other reported measures of personality features and functional skills of the affected patient.

Methods

Study Design

Analysis for the current study focused on identifying baseline characteristics associated with caregiver-reported wandering over time. Both participants and their caregivers provided study data through interviews and self-reported questionnaires. Data were collected through face-to-face interviews with participants and their caregivers at baseline and at 8-month intervals (8, 16, and 24 months). Study data were collected from 2008 to 2012.

All research activities were approved by University of South Florida’s (USF) Institutional Review Board and the James A. Haley VA Research and Development Committee. A waiver for the Health Insurance Portability and Accountability Act enabled study researchers to screen for potential veteran participants by reviewing the Department of Veterans Affairs Computerized Patient Record System (CPRS), Veteran Integrated Service Network Support Services Center (VSSC), and Decision Support System (DSS) databases.

Participants

Veterans were recruited from the James A. Haley Veterans Hospital (VA) and from the USF Memory Disorders Clinic. To be eligible for inclusion in this study, participants had to be (1) 60 years of age or older, (2) newly diagnosed with mild dementia (with “mild” defined as having a Clinical Dementia Rating (CDR) Scale <1 and a Mini-Mental State Examination (MMSE) score >18), (3) living in a home-type setting within 60 miles of the hospital, (4) with a caregiver who was willing and able to serve as a study partner, and (5) independently ambulatory (with or without assistive devices) at point of entry. Both the participant and their caregiver had to speak English. Participants were excluded if they were (1) living in a nursing home or assisted living facility at point of entry; (2) without a designated caregiver living with them; and (3) bilaterally deaf, blind, or had a major psychiatric disorder (excluding temporary depression resulting from dementia).

Veteran participants receiving care from the VA were screened for eligibility (N = 5675) by review of their medical record across 3 VA databases (CPRS, VSSC, and DSS). Additionally, veterans were also referred to study research assistants by clinical staff from the USF Memory Disorders Clinic. Given study inclusion/exclusion criteria, 519 veterans (9.1% of veterans reviewed) were deemed eligible based on the screening process. After eligible veterans were identified, letters describing the study protocol (with an opt-out clause) were mailed to participants with subsequent follow-up telephone calls if no opt-out request was received. Several additional efforts were made to enhance recruitment, including placing flyers on community volunteer bulletin boards, notices in Websites and newsletters as well as conducting presentations to community groups. No data are available regarding the number of veterans referred to study staff at the USF Memory Disorders Clinical and screened, but not recruited.

As a result of these efforts, 143 caregiver dyads (veterans with dementia and their caregiver) consented to participate in the study, with the caregivers serving as informants for personality and behavioral measures; direct assessments of the participants were conducted for cognitive measures. The majority of enrolled participants were recruited from the VA (n = 123; 86.0% of sample), with 20 participants enrolled from USF (14.0% of sample). Baseline demographic characteristics for the complete study sample and differences between those enrolled from the VA and the USF recruitment locations are presented in the Results section.

Measures

Wandering behavior

Revised Algase Wandering Scale–Community Version. ⁸

The Revised Algase Wandering Scale–Community Version (RAWS-CV) is a 40-item tool with 6 subscales to assess wandering behaviors (eloping behaviors, negative outcomes, mealtime impulsivity, persistent walking, repetitive walking, and spatial disorientation), and a total score scale. Previous research on the RAWS-CV indicated the tool to produce valid and reliable scores: test–retest reliability (α = .61-.70), Cronbach’s α for internal consistency for the total score ranging from .87 to .94, and from .83 to .94 for subscales; and discriminant validity values ranging from r = −.20 to −.57. ^8,9 Responses to the RAWS-CV are scored on a 5-point Likert-type scale ranging from Never/Unable (1) to Always (5), for specific wandering behaviors (39 items). Item pool total scores can thus range from 39 to 190; mean item scores are used in analyses. Higher scores indicate higher levels of wandering. One additional item assesses caregiver-reported wandering. Responses for this item ranged from (1) Definitely Not; (2) At Times; (3) Yes, but it is not a problem; and (4) Yes, and it is a problem. Caregiver responses were dichotomized such that “Definitely Not” was categorized as No Wandering (0), with all other responses categorized as Wandering (1), at each study assessment period. For the purposes of these analyses, participants who met this dichotomous criterion for wandering at any point across the study assessment periods were considered to exhibit wandering (1). RAWS-CV data were not available for 1 participant; thus, 1 case is missing from our caregiver-reported wandering analyses. Data for the current study produced scores with moderate to good levels of internal consistency: Eloping Behaviors (α = .77), Negative Outcomes (α = .75), Mealtime Impulsivity (α = .81), Persistent Walking (α = .87), Repetitive Walking (α = .75), Spatial Disorientation (α = .71), and Total Score (α = .90).

Risk factors

Demographics included participant age, gender, race, ethnicity, and education obtained by survey from caregiver report.

Big Five Inventory

The Big Five Inventory was used as a measure of personality in the current study, adapted from the 240-item NEO-PI (Personality Inventory) ¹⁰ and NEO-FFI (Five-Factor Inventory). This measure contains 44 items rated on a Likert-type scale (1 = Disagree Strongly to 5 = agree strongly). The Big Five Inventory was administered to caregivers as a proxy measure of personality in 5 domains: Neuroticism, Extroversion, Openness, Agreeableness, and Conscientiousness. ¹¹ Subscale scores for each domain represent the sum of the items within the domain, with higher scores reflecting the personality trait. Previous research has indicated the Big Five Inventory produces scores with good convergent validity (r = .91) and reliability. ¹¹ In the current study, the Big Five Inventory subscales demonstrated moderate to acceptable levels of reliability: Agreeableness (α = .85), Conscientiousness (α = .86), Extroversion (α = .74), Neuroticism (α = .86), and Openness (α = .77).

Behavioral Response to Stress Scale

The Behavioral Response to Stress Scale (BRSS) ¹² is a 19-item retrospective caregiver questionnaire designed to evaluate the motor responses to stress in persons with dementia who wander. The BRSS assesses 4 domains of behavior: Passivity, Aggressiveness, Motor Busyness, and Negative Verbalization. Caregivers reported the frequency of behaviors related to stressful events for each item, rated on a 5-point Likert-type scale (0 = Never and 4 = Always). Subscale scores for each behavioral domain were calculated as mean scores, with higher scores more likely reflecting the behavioral presentation. Previous research has indicated the BRSS to produce scores with good reliability estimates ranging from .77 to .82. ¹³ In the current study, subscales of the BRSS produced scores with moderate levels of reliability: Aggressiveness (α = .73), Motor Busyness (α = .72), Negative Verbalization (α = .73), Passivity (α = .85), and Total Score (α = .81).

Charlson Co-Morbidity Index

The Charlson Co-Morbidity Index (CC-MI) was developed for use in longitudinal studies as a method for classifying comorbid conditions related to mortality based on retrospective administrative data. ¹⁴ The CC-MI is a weighted index including 19 illnesses. Developed with an initial cohort of medical patients (N = 559), the CC-MI has been shown to be associated with stepwise increases in cumulative mortality attributable to comorbid disease in a second cohort (N = 685) during a 10-year follow-up. ¹⁵

Tinetti Gait and Balance Scale

The Tinetti Gait and Balance Scale was developed to be administered to persons with dementia to assess their gait and balance. ¹⁶ Nine gait components are rated by trained research assistants while participants walk on an even floor (initiation, step height, step length, step symmetry, step continuity, path deviation, trunk stability, walk stance, and turning while walking). Eight balance components are also assessed (sitting balance, arising, immediate standing, standing balance, standing balance with eyes closed, nudged, turning 360 degrees, and sitting down). Scores for each component are rated on a 3-point scale, ranging from 0 (most impairment) to 2 (independence). Individual scores are then aggregated to form two balance and gait subscales as well as a balance and gait total score. Previous research on this scale produced scores demonstrating excellent test–retest reliability (r = .72-.93) with good discriminant validity. ¹⁷ For the current study, subscales of the Tinetti Gait and Balance produced scores with high levels of reliability: balance (α = .91), gait (α = .90), and balance and gait total (α = .94).

Modified scale for Instrumental Activities of Daily Living

The Instrumental Activities of Daily Living (IADLs) is an 8-category scale assessing housework, shopping, cooking, laundry, telephone, medication, and transportation. ¹⁸ The scale reflects gender differences inherent in IADL behaviors by using different scores for tasks considered primarily male or female by people now aged over 70. A higher total score reflects higher independence in daily living skills. In previous research, the modified scale displayed high internal consistency among a sample of persons with dementia ( α = .82). ¹⁹ A total score for the IADLs demonstrated moderate levels of internal consistency in the current study (α = .77).

The CDR Scale

The CDR Scale is an instrument for dementia staging, ²⁰ utilizing caregiver ratings for 6 categories of functioning (Memory, Orientation, Judgment and Problem-Solving, Community Affairs, Home and Hobbies, and Personal Care). Caregivers provide ratings on a 5-point Likert-type scale, with lower scores indicating better function. Subscales are produced categorically and globally. The CDR has been used widely in longitudinal studies and clinical trials for staging severity of dementia. ²¹ Previous research utilizing the CDR indicates that the CDR produces scores with strong convergent validity with cognitive measures ²² as well as acceptable levels of inter-rater reliability (κ = .58-.85). ^23,24 For the current study, CDR scores were obtained through caregiver report.

The MMSE

The MMSE ²⁵ is a widely used, brief, and standardized method for assessing cognitive mental status. The MMSE consists of 11 items that assess orientation, attention, immediate and short-term recall, language, and the ability to follow simple verbal and written commands. Items are weighted and summed to yield a total score, ranging from 0 to 30, with higher scores indicating a better cognitive mental status. Previous research indicates the MMSE produces scores with good levels of reliability and validity as well as sensitivity (0.86) and specificity (0.92). ²⁵

Data collection procedures

Data on participant wandering behavior and the potential risk factors discussed above were collected at baseline during scheduled clinic or home visits. Caregivers completed study assessments at baseline and subsequent follow-up intervals (8, 16, and 24 months after baseline assessment). Study measures were collected at in-person visits involving the participant and caregiver. Study data were collected by 2 trained research assistants and the project manager. Research assistants and the project manager received uniform training in data collection procedures to support consistency of study data collection.

Analyses

Frequencies and descriptive statistics were used to portray the baseline characteristics of the study sample, RAWS-CV data (subscales, total score, and caregiver-reported wandering item He/She is a wanderer) as well as caregiver-reported wandering across study visits. A series of χ² and t test analyses were used to examine potential differences between veterans who were recruited from the VA compared to those from the USF Memory Disorders Clinic. An additional series of χ² and t test analyses were used to explore group differences in baseline characteristics between (1) those with and without caregiver-reported wandering as well as (2) those who did and did not demonstrate inconsistency in caregiver-reported wandering status across study visits (ie, described as demonstrating wandering at 1 visit but not a subsequent visit). Further, a series of bivariate analyses were used to examine potential differences between those with full versus partial study data, to explore potential impact of sample attrition on the generalizability of study findings. Bonferroni family-wise error corrections were applied by dividing the nominal P value of .05 by the number of statistical tests conducted within a family of predictor variables and to interpret the more conservative P value. For example, for a family of tests with 3 scales, the P value would be .017 (.05/3).

An exploratory multivariable logistic regression model was conducted to predict caregiver-reported wandering (no, yes). This model utilized variables significant in bivariate analyses after application of the Bonferroni family-wise error correction, excluding the Balance subscale of the Tinetti Gait and Balance measure. This subscale was excluded due to multicollinearity between the Total scale of the Tinetti Gait and Balance measure, with observed variance inflation factor (VIF) values above 10.0 (for both parameters).

All tests were two-tailed and evaluated at the P < .05 level. Additionally, results presented are for robust test statistics, accounting for any observed inequality of variances between groups assessed for differences.

Results

Sample Demographics, Baseline Characteristics, and Differences by Enrollment Location

Table 1 presents the demographic characteristics of the study sample (N = 143). Baseline characteristics for study measures are also presented for the entire sample in Table 1; study measures by caregiver-reported wandering and other group variables are discussed in further detail below.

Table 1.

Participant Demographics and Baseline Characteristics.^a

	Total(N = 143)
Variable	n	M (SD) or %
Demographics
Age	–	77.50 (7.16)
Gender	–	–
Male	129	90.2
Female	14	9.8
Race	–	–
American Indian or Alaska Native	0	0.0
Asian or Pacific Islander	0	0.0
Black	5	3.5
White	134	94.4
Other	3	2.1
Ethnicity	–	–
Hispanic	9	6.3
Non-Hispanic	134	93.7
Degree (education)	–	–
No formal school completed	0	0.0
Less than 9th grade	6	4.4
9th through 12th grade but did not finish high school	16	11.7
High school graduate or GED	40	29.2
Some college but no degree	31	22.6
Associate degree	12	8.8
Bachelor’s degree	13	9.5
Master’s degree	15	10.9
PhD	3	2.2
Other	1	0.7
The Big Five Inventory: NEO Five-Factor Personality Inventory
Personality: Agreeableness	–	34.01 (8.30)
Personality: Conscientiousness	–	28.51 (8.70)
Personality: Extroversion	–	24.64 (6.66)
Personality: Neuroticism	–	25.77 (7.93)
Personality: Openness	–	29.12 (7.54)
Behavioral Response to Stress Scales (BRSS)
Aggressiveness	–	0.40 (0.47)
Motor busyness	–	1.18 (0.83)
Negative verbalization	–	0.82 (0.77)
Passivity	–	1.23 (0.95)
Charlson Comorbidity Index (CC-MI)
Class 1	–	1.72 (0.86)
Class 2	–	0.61 (1.27)
Class 3	–	0.00 (0.00)
Class 4	–	0.00 (0.00)
Health status (based on subscales)	–	2.34 (1.63)
Tinetti Gait and Balance Scale
Mobility: Balance	–	11.71 (3.83)
Mobility: Gait	–	9.21 (3.31)
Mobility: Balance and Gait Total	–	20.92 (6.88)
Modified scale for Instrumental Activities of Daily Living (IADLs)
Global score	–	4.61 (2.09)
Mini-Mental State Examination (MMSE)
Mini-Mental State Examination—Total score	–	24.69 (3.50)
RAWS-CV
Eloping behavior	–	1.25 (0.44)
Negative outcome	–	1.42 (0.61)
Mealtime impulsivity	–	1.21 (0.60)
Persistent walking	–	1.41 (0.50)
Repetitive walking	–	1.51 (0.61)
Spatial disorientation	–	1.20 (0.49)
RAWS total	–	1.37 (0.37)

Abbreviations: GED, general equivalency degree; RAWS, Revised Algase Wandering Scale; M: Mean; SD, standard deviation.

^aSome missing data; valid percentages displayed.

The average age of participants was 77.50 years (standard deviation [SD] = 7.16), with a range from 61 to 92 years. Most participants were male (90.2%), white (94.4%), and non-Hispanic (93.7%). Participants most commonly completed some level of college education (31.4%), a high school degree or their general equivalency degree (GED; 29.2%), followed by some college but with no degree (22.6%). A minority of participants (16.1%) had less than a high school education.

There were no statistically significant demographic differences between participants recruited from the VA (n = 123) and those recruited from USF (n = 20), after application of Bonferroni family-wise error correction. With the exception of the Tinetti Gait and Balance Scale, there were no statistically significant differences between participants at the VA and USF enrollment locations in clinical measures assessed at baseline. Across both subscales and total score of the Tinetti Gait and Balance Scale, participants from USF had higher scores compared to VA participants: (1) Balance subscale (M = 13.79 [SD = 2.64] compared to M = 11.38 [SD = 3.89], t(31.97) = −3.43, P = .002), (2) Gait subscale (M = 11.00 [SD = 1.83] compared to M = 8.93 [SD = 3.41], t(41.44) = −3.98, P < .001) as well as (3) the Balance and Gait Total Scale (M = 24.79 [SD = 4.20] compared to M = 20.31 [SD = 7.03], t(36.44) = −3.87, P < .001). All analyses were still significant after application of a Bonferroni family-wise error correction (P < .05/3 [.017]).

RAWS-CV Subscale, Total Score, and Caregiver-Reported Wandering Data Across Study Visits

Table 2 presents participant RAWS-CV data across study visits (subscales, total score, and caregiver-reported wandering item). Across all subscales, total score, and caregiver-reported wandering item (He/She is a wanderer), participants scored very low in caregiver-reported wandering behaviors, with scores averaging between Never and Seldom. Additionally, these low scores did not vary substantially in severity across all study visits, indicating a very low level of intensity of wandering behaviors (as measured by the RAWS-CV), in our cohort of veterans with early dementia.

Table 2.

RAWS-CV Subscale, Total Score, and Caregiver-Report of Wandering by Study Visit^a.

	Visit 1 (Baseline; n = 142)				Visit 2 (8-month Follow-Up; n = 119)				Visit 3 (16-month Follow-Up; n = 100)				Visit 4 (24-month Follow-Up; n = 98)
RAWS-CV Scale/Item	M	SD	Min	Max	M	SD	Min	Max	M	SD	Min	Max	M	SD	Min	Max
Subscales
Eloping behavior	1.25	0.44	1.00	3.17	1.23	0.41	1.00	3.17	1.29	0.51	1.00	3.67	1.26	0.48	1.00	3.33
Negative outcomes	1.42	0.61	1.00	3.75	1.44	0.64	1.00	4.25	1.43	0.57	1.00	3.50	1.46	0.63	1.00	3.50
Mealtime impulsivity	1.21	030	1.00	5.00	1.12	0.38	1.00	3.50	1.28	0.64	1.00	5.00	1.16	0.36	1.00	3.00
Persistent walking	1.41	0.50	1.00	3.25	1.46	0.50	1.00	3.25	1.46	0.54	1.00	2.83	1.31	0.49	1.00	2.83
Repetitive walking	1.51	0.61	1.00	3.57	1.42	0.56	1.00	3.29	1.46	0.58	1.00	3.14	1.31	0.50	1.00	3.57
Spatial disorientation	1.20	0.49	1.00	3.50	1.25	0.60	1.00	4.00	1.18	0.45	1.00	3.00	1.27	0.55	1.00	3.25
Total score	1.37	0.37	1.00	2.62	1.36	0.37	1.00	2.67	1.38	0.43	1.00	2.67	1.30	0.40	1.00	2.69
He/she is a wanderer	1.18	0.47	1.00	3.00	1.21	0.53	1.00	3.00	1.19	0.51	1.00	3.00	1.17	0.43	1.00	3.00

Abbreviations: RAWS-CV, Revised Algase Wandering Scale–Community Version; M, Mean; SD, standard deviation.

^aSome missing data in RAWS-CV scores and computed indices. Valid data displayed.

Prevalence, Incidence, and Consistency of Caregiver-Reported Wandering Across Study Visits

Based upon our dichotomized wandering variable computed, roughly one-quarter of participants (n = 37; 26.1%) were reported by their caregivers to demonstrate wandering, at 1 or more time point, across all study visits. Of those with caregiver-reported wandering, 17 participants were consistent in caregiver report of wandering (45.9%) across all visits; 20 participants (54.1%) did not appear to be consistently characterized as wanderers, such that they were reported to have wandering by caregivers at 1 visit and then at least 1 subsequent visit where they were not reported to wander.

Table 3 presents the proportion of caregiver-reported wandering based on the global measure of reported wandering, across study visits as well as the consistency of wandering behavior across study visits. Although there was variation across all time points, note that roughly 14% to 15% of participants were reported by caregivers to wander at each assessment point, across all study visits.

Table 3.

Caregiver-Reported Participant Wandering^c, and Consistency Across Visits.^a,b

	Visit 1 (n = 139)		Visit 2 (n = 117)		Visit 3 (n = 100)		Visit 4 (n = 98)
Caregiver-Reported Wandering Status	n	%	n	%	n	%	n	%
No wandering	119	85.6	100	85.5	86	86.0	83	84.7
Wandering	20	14.4	17	14.5	14	14.0	15	15.3
Consistent wandering	7	5.0	8	6.8	8	8.0	9	9.2
Inconsistent wandering	13	9.4	9	7.7	6	6.0	6	6.1

Abbreviation: RAWS-CV, Revised Algase Wandering Scale–Community Version.

^aN = 37.

^bSome missing data in RAWS-CV scores (and computed caregiver-reported wandering variable) across study assessments; valid percentages displayed.

^cCaregiver-reported wandering was defined as any caregiver positive endorsement to the item He/she is wanderer, across 1 or more of study visits observed.

Exploring Baseline Differences in Risk Factors by Caregiver-Reported Wandering

Independent t test and χ² analyses demonstrated some statistically significant differences between those participants with (n = 37; 26.1%) and those without (n = 105; 73.9%) caregiver-reported wandering at baseline. Results are presented in Table 4 and are discussed separately by each measure below.

Table 4.

Examining Group Differences in Baseline Variables Between Those With and Without Caregiver-Reported Wandering.^a,b

	No Wandering (n = 105)		Wandering (n = 37)		Significance
Subscale	n	M (SD) or %	n	M (SD) or %	Test statistic	Bonferroni adjustment
Demographics
Age	–	77.81 (7.29)	–	76.86 (6.76)	t(140) =0.69, P = .491	NS
Gender	–	–	–	–	χ2(1) = 1.12, P = .291	NS
Male	93	88.6	35	94.6	–	–
Female	12	11.4	2	5.4	–	–
Race	–	–	–	–	χ2(2) = 2.79, P = .248	NS
American Indian or Alaska Native	0	0.0	0	0.0	–	–
Asian or Pacific Islander	0	0.0	0	0.0	–	–
Black	4	3.8	1	2.8	–	–
White	100	95.2	33	91.7	–	–
Other	1	1.0	2	5.6	–	–
Ethnicity	–	–	–	–	χ2(1) = 4.34, P = .037	NS
Hispanic	101	96.2	32	86.5	–	–
Non-Hispanic	4	3.8	5	13.5	–	–
Degree (education)	–	–	–	–	χ2(8) = 8.41, P = .395	NS
No formal school completed	0	0.0	0	0.0	–	–
Less than 9th grade	4	4.0	2	5.7	–	–
9th through 12th grade but did not finish high school	9	8.9	7	20.0	–	–
High school graduate or GED	30	29.7	9	25.7	–	–
Some college but no degree	24	23.8	7	20.0	–	–
Associate degree	7	6.9	5	14.3	–	–
Bachelor’s degree	12	11.9	1	2.9	–	–
Master’s degree	11	10.9	4	11.4	–	–
PhD	3	3.0	0	0.0	–	–
Other	1	1.0	0	0.0	–	–
The Big Five Inventory: NEO Five-Factor Personality Inventory
Personality: Agreeableness	–	34.88 (8.14)	–	31.54 (8.37)	t(139) = 2.13, P = .035	NS
Personality: Conscientiousness	–	29.80 (8.03)	–	24.89 (9.58)	t(139) = 3.03, P = .003	Sig.
Personality: Extroversion	–	24.76 (6.94)	–	24.30 (5.86)	t(139) = 0.36, P = .718	NS
Personality: Neuroticism	–	24.80 (7.90)	–	28.49 (7.48)	t(139) = −2.47, P = .015	NS
Personality: Openness	–	29.20 (7.07)	–	28.89 (8.83)	t(139) = 0.21, P = .831	NS
Behavioral Response to Stress Scales (BRSS)
Aggressiveness	–	0.33 (0.39)	–	0.62 (0.58)	t(48.30) = −2.84, P = .007	Sig.
Motor busyness	–	1.18 (0.87)	–	1.22 (0.74)	t(137) = −0.24, P = .814	NS
Negative verbalization	–	0.72 (0.78)	–	1.10 (0.66)	t(137) = −2.62, P = .010	Sig.
Passivity	–	1.10 (0.93)	–	1.61 (0.92)	t(137) = −2.84, P = .005	Sig.
Charlson Co-Morbidity Index (CC-MI)
Class 1	–	1.73 (0.87)	–	1.65 (0.75)	t(137) = 0.48, P =.635	NS
Class 2	–	0.63 (1.35)	–	0.59 (1.04)	t(137) = 0.13, P = .893	NS
Class 3	–	0.00 (0.00)	–	0.00 (0.00)	c	c
Class 4	–	0.00 (0.00)	–	0.00 (0.00)	c	c
Health Status (based on subscales)	–	2.35 (1.74)	–	2.24 (1.28)	t(137) = 0.35, P = .727	NS
Tinetti Gait and Balance Scale
Mobility: Balance	–	12.23 (3.80)	–	10.30 (3.57)	t(137) = 2.68, P = .008	Sig.
Mobility: Gait	–	9.56 (3.09)	–	8.24 (3.74)	t(137) = 2.10, P = .038	NS
Mobility: Balance and Gait Total	–	21.78 (6.68)	–	18.54 (6.95)	t(137) = 2.51, P = .013	Sig.
Modified scale for Instrumental Activities of Daily Living (IADLs)
Global score	–	4.89 (2.10)	–	3.92 (1.85)	t(139) = 2.50, P = .014	Sig.
Mini-Mental State Examination (MMSE)
Mini-Mental State Examination—total score	–	24.79 (3.69)	–	24.57 (2.83)	t(81.75) = 0.38, P = .706	NS
RAWS-CV
Eloping behavior	–	1.14 (0.25)	–	1.55 (0.64)	t(40.03) = −3.83, P < .001	Sig.
Negative outcome	–	1.32 (0.51)	–	1.72 (0.75)	t(48.20) = −3.01, P = .004	Sig.
Mealtime impulsivity	–	1.16 (0.57)	–	1.37 (0.68)	t(54.59) = −1.66 P = .103	NS
Persistent walking	–	1.34 (0.46)	–	1.63 (0.57)	t(53.50) = −2.88, P = .006	Sig.
Repetitive walking	–	1.41 (0.55)	–	1.75 (0.62)	t(140) = −3.06, P = .003	Sig.
Spatial disorientation	–	1.13 (0.40)	–	1.37 (0.66)	t(45.57) = −2.10, P = .041	NS
RAWS total	–	1.28 (0.28)	–	1.61 (0.45)	t(46.52) = −4.31, P < .001	Sig.

Abbreviations: GED, general equivalency degree; NS = not significant; RAWS, Revised Algase Wandering Scale; Sig., significant; M, Mean; SD, standard deviation.

^aN = 142_.

^bSome missing data; valid data displayed.

^cCould not compute due to lack of variance.

Demographics

There were no statistically significant differences in baseline demographic characteristics between those with and without caregiver-reported wandering.

The Big Five Inventory: NEO Five-Factor Personality Inventory

Statistically significant differences were observed between those with and without caregiver-reported wandering in the Conscientiousness subscale (M = 24.89 [SD = 9.58] compared to M = 29.80 [SD = 8.03] for nonwanderers). There were no other statistically significant differences between those with and without caregiver-reported wandering in NEO Five-Factor Personality Inventory subscales.

The BRSS

Participants with caregiver-reported wandering had higher scores on their baseline Negative Verbalization subscale scores (M = 1.10 [SD = 0.66] compared to M = 0.72 [SD = 0.78]). Participants with caregiver-reported wandering had higher scores on their baseline Passivity subscale scores (M = 1.61 [SD = 0.92] compared to M = 1.10 [SD = 0.93]). Additionally, participants with caregiver-reported wandering had higher score on their baseline Aggressiveness subscale scores (M = 0.62 [SD = 0.58] compared to M = 0.33 [SD = 0.39]). There were no significant differences observed between those with and without caregiver-reported wandering for the Motor Busyness subscales of the BRSS.

The CC-MI

There were no statistically significant differences observed between those with and without caregiver-reported wandering in baseline CC-MI subscales.

Tinetti Gait and Balance Scale

Participants with caregiver-reported wandering had lower scores on their baseline Balance subscale scores (M = 10.30 [SD = 3.57] compared to M = 12.23 [SD = 3.80]). Additionally, participants with caregiver-reported wandering had lower scales on the Balance and Gait Total Scale scores (M = 18.54 [SD = 6.95] compared to M = 21.78 [SD = 6.68]), indicating greater degree of balance instability.

Modified scale for IADLs

Participants with caregiver-reported wandering had lower scores on their baseline Global Score for the IADLs (M = 3.92 [SD = 1.85] compared to M = 4.89 [SD = 2.10]), indicating a greater need for daily assistance.

The MMSE

There were no statistically significant differences in the MMSE Total Scores between those with and without caregiver-reported wandering.

Exploring Baseline Differences in Wandering Measures by Caregiver-Reported Wandering

The RAWS-CV

For most subscales and the total score of the RAWS-CV (excluding the Mealtime Impulsivity and Spatial Disorientation subscales), statistically significant differences were observed between those with and without caregiver-reported wandering. Results indicate higher scores on most RAWS-CV subscales and total score (indicating a higher degree of wandering intensity), among those with caregiver-reported wandering. These statistically significant findings remained even after application of the Bonferroni family-wise error correction (p = .05/8 = .00625).

Baseline Differences Between Those With and Without Consistency in Caregiver-Reported Wandering Status

An additional series of independent t test and χ² analyses were conducted to see if those who were consistent (n = 17) versus inconsistent in their caregiver-reported wandering status (n = 20) across study visits differed in their baseline scores. With the exception of the IADLs, there were no statistically significant baseline differences observed between those who did (n = 20; 54.1%) and did not vary (n = 17; 45.9%) in their caregiver report of wandering. Participants with variance in caregiver-reported wandering had higher scores on the IADLs (indicated less severe problems and higher levels of functioning) compared to those without variance (M = 4.45 [SD = 2.11], vs M = 3.29 [SD = 1.26]), t(31.67) = −2.05, P = .049).

Examining Differences in Baseline Characteristics Among Those With and Without Study Attrition

Across the course of this community-based longitudinal study, attrition was observed for roughly a third of our sample (n = 50; 35.0%). Most participants completed all 4 study assessments (n = 93; 65.0%), with some participants completing 3 visits (n = 12; 8.4%), some 2 visits (n = 16; 11.2%), and with some only completing the baseline assessment (n = 22; 15.4%). Exploratory bivariate analyses were conducted to see if there were any baseline differences between those with full (65.0%) and partial (35.0%) study data.

No significant differences were observed in measured baseline characteristics between those with and without full study data, after application of the Bonferroni family-wise error correction. Further, those with complete versus partial study data did not statistically significantly differ in their rate of any caregiver-reported wandering over the study period.

Exploratory Logistic Regression Model Predicting Caregiver-Reported Wandering

An exploratory multivariable logistic regression model was conducted to predict those with caregiver-reported wandering, using variables significant in bivariate analyses after application of the Bonferroni family-wise error correction. For this modeling procedure, only the Balance and Gait Total Score of the Tinetti Gait and Balance Scale was used, with the Balance score excluded (due to concerns with multicollinearity; VIF values of >10.0 observed for both parameters). Other measures of wandering behavior (eg, RAWS-CV Total Score) were not included in the logistic regression model. Although a significant model was achieved, χ²(6) = 20.12, P = .003, Hosmer-Lemeshow goodness of fit χ²(8) = 3.62, P = .890, there were no statistically significant model parameters.

Discussion

This study represents one of the first studies to longitudinally examine wandering behavior in persons with early dementia in a community-based setting. To date, most studies of wandering behaviors have been focused on those with more advanced dementia and/or those residing in residential care settings. Relative to the many studies that have addressed wandering in long-term care setting, only a very small percentage have examined wandering behaviors longitudinally among those with mild dementia residing in the community. As such, this study provides preliminary evidence regarding the baseline clinical characteristics of those with the presence of any caregiver-reported wandering behaviors in early dementia. Understanding these characteristics and how families report them may aid in the development of family-focused interventions that may serve to help persons with early dementia remain in the community and avoid or postpone long-term care placement.

This study attempted to provide a rough estimate of the incidence and prevalence of wandering in a sample of community-dwelling veterans with early dementia. For these purposes, the RAWS-CV item for caregiver-reported wandering was dichotomized (No Wandering, Any Wandering), with any indication of wandering at 1 or more study visits, used to classify caregiver-reported wanderers. Across 2 years of community-based observation in this study, roughly one-quarter of participants (n = 37; 26.1%) were reported by their caregivers to demonstrate wandering behavior at 1 or more point across all study visits. Although there was some variability in caregiver report of participant wandering across study visits, roughly 15% of participants at each study visit were reported to wander. Results from the continuous subscales and total score of the RAWS-CV indicated a very low level of intensity of wandering behaviors. Furthermore, the subscale and total scores for the RAWS-CV were relatively stable in severity across the 2 years of study observation. This is not surprising when considering that behavioral symptoms in dementia may fluctuate in frequency over time but may not always fluctuate in severity. This may lead to symptoms being reported inconsistently, as observed here, but when they do occur, the severity may not necessarily differ.

These findings may appear to contradict previous research indicating that wandering increases as the severity of dementia increases. Devanand and colleagues ²⁶ examined the prevalence of wandering over time with participants in a naturalistic observation study over 5 years. Results from Devanand et al ²⁶ found that 38.7% of participants demonstrated wandering, with this prevalence increasing to 56.9% by the third year of observation. These prevalence figures contrast sharply with the results from the current study, with roughly 15% of participants with caregiver-reported wandering, at each assessment period. This prevalence rate was stable across 2 years of observation (with sample attrition). However, our sample only included individuals who were newly diagnosed with dementia, as compared to those from Devanand et al ²⁶ who had an average of 4 years since their dementia diagnosis at the time of study enrollment. Findings from the current study suggest that wandering may not necessarily be more prevalent into several years into a persons’ diagnosis of dementia. More work is needed to understand both the persistence and progression of wandering as individuals with dementia progress in their disease, and how these measures may be ameliorated through behavioral interventions.

The global measure of wandering used in the current study, created by the dichotomization of the item, He/She is a wanderer, provided a method for the longitudinal detection of the behavior to be compared to baseline clinical features. This resulted in our findings of an association of wandering with lower baseline performance in IADL function and worse performance in gait and balance measures. Further, the interesting association of specific personality characteristics with wandering may help inform future studies that test family-based interventions. In addition—and while expected—the significant differences in most RAWS-CV subscales and total scores between those with and without caregiver-reported wandering provide some evidence for the convergent validity of caregiver-reported wandering.

Further research is needed to examine the validity and utility of caregiver-reported wandering for research and clinical purposes. It should also be noted that our study operationalized caregiver-reported wandering as any reported wandering in our two years of community-based observation; and there was variation in the consistency of caregiver-reported over time. Findings and associations observed may therefore differ in those with more consistent and/or higher levels of wandering (as reported by caregivers or other measures of wandering). It is also possible that behavioral fluctuations inherent to the illness contribute to variance in the behavior. More work is needed to more precisely examine the emergence, continuation, and exacerbation of wandering behaviors in persons with early dementia as they progress through their disease.

With that caveat, exploratory bivariate analyses identified several baseline characteristics that distinguished the patients characterized by caregivers to wander at some point in the study, with a subset of differences remaining statistically significant after correcting for the Bonferroni family-wise error rate. Those with caregiver-reported wandering were also found to have lower scores on the conscientiousness subscale of the NEO Big Five Personality Inventory) as well as higher scores on 3 subscales of the BRSS (aggressiveness, negative verbalization, and passivity). An exploratory logistic regression model was conducted to predict caregiver-reported wandering, using those variables significant in bivariate analyses (after application of Bonferroni family-wise error correction). Although a significant model was reached, all model parameters failed to reach significance. Future research is needed to examine whether these baseline characteristics may indeed be useful in predicting caregiver-reported wandering, whether additional measures of wandering may be fruitful for such examination, and/or whether these associations may be evident with a larger sample with greater power to detect effects.

Findings from the current analysis support and expand upon a previous study published with data from this study cohort. ²⁷ Ali and colleagues ²⁷ examined predictors associated with any caregiver endorsement of items contained within the negative outcomes and eloping behavior subscales of the RAWS-CV at baseline, utilizing separate logistic regression models. Their analyses indicated associations between the RAWS-CV persistent walking subscale and both dichotomous outcomes as well as associations between the Passivity subscales of the BRSS and any elopement behavior, and the Total Gait and Balance subscale of the Tinetti Gait and Balance scale and any negative outcomes at baseline. Our study builds upon the previous results of Ali et al, ²⁷ with a longitudinal assessment of wandering behavior; our results also indicate statistically significant association with wandering and BRSS measures.

Additionally, our results support findings from two other studies that also examined factors associated with wandering behaviors. ^12,28 Schonfeld and colleagues ²⁸ utilized a cross-sectional design with data from 134 VA nursing home facilities, among male veterans with moderate or severe cognitive impairment. Results from their study also indicated that wandering behavior was associated with poorer capacities to conduct basic activities of daily living (ADLs; personal hygiene). The association with wandering and impairment of these basic ADLs is consistent with the more cognitively impaired sample studied by Schonfeld and colleagues. Although the IADLs measures reflect higher functioning skills related to ADLs, our finding that wandering behavior was associated with impairment in ADLs is consistent with their previous research.

Song and Algase ¹² also employed a cross-sectional design, among a largely female population of persons in long-term care facilities. This study also utilized many of the same measures employed in the current study. Results from their analyses also indicated that the Negative Verbalization subscale was associated with wandering behavior, even among a sample of persons with much higher levels of dementia (as evidenced by poor cognitive functioning as measured by their MMSE scores, as well as placement in long-term care facilities).

Limitations

Findings in the current study may be tempered by our sample of only veterans who were predominantly white and male, which may limit its representativeness relative to the diverse population of older adults who will experience dementia. Caregiver-reported wandering in our analyses may also be underreported due to sample attrition and missing data in later follow-up assessments among patients with poor outcomes who were not retained in follow-up. It is not clear whether our sample attrition was due to advancements in participants’ disease states, participant transfer from home-based to long-term care facilities, logistical issues, or other possibilities. However, exploratory analyses examining baseline differences between those with and without study attrition failed to observe any significant differences, suggesting the potential generalizability of study findings. There may also be measurement concerns regarding the self-reported nature of the dependent variable. Further, we operationalized caregiver-reported wandering as any report of participant wandering in the 2 years of study observation, with some participants found to vary in caregiver-reported wandering. Our findings should thus most conservatively be considered to reflect the characteristics of persons newly diagnosed with dementia who display some wandering. These characteristics may differ among those with more persistent or higher levels of wandering. Additionally, our study was underpowered to examine predictors of caregiver-reported wandering over time; future studies with larger sample sizes should examine longitudinal predictors of wandering over time.

Implications for Practice and Future Research

This study represents one of the first community-based longitudinal studies of those with early dementia, documenting the incidence and prevalence of caregiver-reported wandering behaviors as well as exploring baseline characteristics that may be associated with these observed wandering behaviors. Although focusing on veterans, this study nonetheless presents preliminary evidence of characteristics among those with early dementia that may be amenable to early intervention efforts. Specifically, this study found that patients with caregiver-reported wandering exhibited lower scores on certain personality aspects (conscientiousness) as well as higher, poorer scores on certain behavioral responses to stress (Aggressiveness, Negative Verbalization, and Passivity). Further, the findings of greater daily living impairment and poorer performance in gait and balance tasks suggest that there was a higher degree of overall decline in the group characterized as exhibiting wandering behaviors.

The inconsistent reporting by caregivers of wandering behavior was an unexpected finding that may have been derived from a variety of sources. One of these is the fluctuating nature of behavioral symptoms in dementia. For symptoms that occur infrequently or at a low level, caregivers may not perceive or report them as a particular problem, even when objective reporting or ratings suggest the behavior is present. Further, given that wandering may appear goal directed, the behavior may be interpreted or rationalized by caregivers as representing purposeful activity as opposed to a symptom of dementia. To enhance safety for patients with dementia, a better understanding of this finding may warrant additional study.

Additionally, this raises methodological questions of the ability of caregivers to report wandering when exhibited at low levels of intensity. This could impact the ability to recruit participants with wandering behaviors for future studies and could be of particular concern when developing interventional research for those with mild dementia who wander. For the current study, however, we were interested only in providing a broad examination of any wandering behavior among those with early dementia and potential associations with baseline characteristics. This broad dichotomization, across 2 years of observation and 4 assessment periods, provided some important preliminary findings of factors associated with wandering among a community-based sample of persons with early dementia. More work is needed to more precisely understand if and how caregivers may perceive wandering at a low level of intensity and frequency and how this may be examined and utilized in future research efforts.

Individuals with dementia receiving care at home may be in a unique environment that is responsive to tailored interventions for a family-care setting that may permit continued living at home for a longer duration. For example, it is known that motor memory consolidation may increase the likelihood of continued performance of highly consolidated motor tasks in familiar settings. These “overlearned motor tasks” may include playing the piano or other complex motor behaviors including dancing, driving, and so on. Highly consolidated memory tasks may include walking in familiar environment such as one’s community home or neighborhood of many years where the likelihood of becoming lost is reduced by the overlearned behavior, much like driving a familiar route. Given that the home environment may lend itself to walking routes that are relatively preserved in memory, an individual with restless motor behavior may be able to move safely around a home environment and avoid the hazards of memory interference induced by an unfamiliar environment. ²⁹ Consequently, living in one’s home may be an environment that is particularly suitable for interventions that permit “safe” movement monitored by a trained family caregiver.

This study also provides suggestive evidence for personality characteristics and behavioral responses to stress that may be useful to identify and, potentially, intervene in those with early dementia to minimize adverse outcomes associated with wandering (eg, elopement, injurious falls). The observation of low ratings on the conscientiousness subscale in patients reported to wander is particularly intriguing. This scale includes items of perceived “dutifulness, self-discipline, and deliberation” such that low ratings suggest that the individual is perceived as someone likely to behave without concern for order and duty. This may be useful to explore in a family intervention setting, such that families may associate wandering as related to a careless personality, which may induce more feelings of distress in the caregiver due to the perception that there is a volitional disregard for order in the household. This is supported by the observation that family members may tend to perceive greater functional skills as retained in individuals with dementia when they are not supported by objective testing. ³⁰

Addressing this possible factor with caregivers may lead to successful strategies to prevent the caregiver distress that often contributes to long-term care placement. As Brodaty and Arasaratnam ⁷ noted, there is evidence in the family care setting for enhanced success of behavioral interventions. Given that keeping persons with dementia at home in the community is exceedingly preferable and in line with quality of life and reducing the cost of care, future intervention work is likely to be of great value.