Elements of Mobility as Predictors of Survival in Elderly Patients with Dementia: Findings from the Canadian Study of Health and Aging

Athanasios Tom Koutsavlis and Christina Wolfson

Abstract

In order to identify elements of mobility that predict survival in elderly people with dementia, we conducted a two-year follow-up of a cohort of dementia subjects from the population-based Canadian Study of Health and Aging. There were 749 prevalent cases of Alzheimer's disease and 208 prevalent cases of vascular dementia. Elements of mobility that predicted death during the two-year follow-up period included difficulty in dressing (OR = 2.08, 95% CI: 1.41-3.07), difficulty in getting about (OR = 1.69, 95% CI: 1.18-2.40), history of falls (OR = 1.43, 95% CI: 1.05-1.94), abnormal gait (OR = 1.61, 95% CI: 1.08-2.40) and abnormal motor strength (OR = 1.51, 95% CI: 1.07-2.15). Sociodemographic factors such as older age and male sex were also significant predictors of decreased survival. These associations are potentially useful to clinicians and health professionals by providing prognostic information to supply to families and suggesting areas in which interventions to improve survival might be focused.

Key words: dementia; mobility; predictors; survival

Introduction

Old age is all too commonly a time of impaired neurologic function. In 1978, the annual cost of caring for patients with dementia in the United States, where dementia is the major debilitating condition of more than half the residents of nursing homes, was $12 billion, and the projected cost by the year 2030 is $30 billion (1978 dollars).¹ In Canada, prevalence estimates in 1991 suggested that 252,600 (8.0%) of all Canadians aged 65 and over met the criteria for dementia.² If the prevalence estimates remain constant, then the number of Canadians with dementia will rise to 592,000 by the year 2021.² In a 1998 study by Hux et al., the annual societal cost of care per patient with Alzheimer's disease (AD) in Canada ranged from $9,451 for mild disease to $36,794 (1996 Canadian dollars) for severe disease.³

Various institution- and community-based studies have indicated that survival is significantly reduced by AD and vascular dementia (VaD).^4-6 At present, the progressive course of dementia cannot be reversed,^7,8 and prognostic data are therefore important in the management of this condition and in the allocation of resources.⁹ Several prognostic factors have been investigated: increasing age,^9-13 male sex,^6,9,10,13-15 functional disability (mobility),^6,8-10,16-19 severity of dementia,^{4,5,8-11,14,15} age of onset,^8,9,11,12,16 educational level,^{9,10,12,14,15} extrapyramidal signs (EPS)^11,20-22 and comorbidity.^1,5,8,14 Only age, sex and mobility have been shown consistently in the literature to be related to shorter survival, and since only mobility is modifiable, it is a vital prognostic factor in dementia.

Poor mobility is postulated to affect survival both by increasing the risk of falls and through secondary diseases related to immobility.⁸ Patients with AD are known to be at considerable risk for falls and fractures compared with non-demented persons of the same age. Fractures tend to immobilize patients even further. Immobility then places them at risk for other potentially lethal events, such as pulmonary embolus and aspiration pneumonia.⁸

Unfortunately, past studies that have identified the value of mobility as a predictor of survival in dementia are hampered by methodologic complications (Table 1).^6,8-10,16-19 Many study populations consisted of subjects from geriatric and dementia clinics, psychiatric hospitals and nursing homes,^8,16,17,19 often with over-representation of advanced cases. The aim of all previous studies has been to detect predictors of disease prognosis,^6,8-10,16-19 and to our knowledge none has concentrated exclusively on investigating mobility. Most studies have restricted their scope to Alzheimer's disease,^8,10,16-18 and only one recent study has investigated all dementias.⁹

Finally, and most importantly, the definition of mobility has been poorly captured in almost all studies. A wide range of proxies for mobility, such as the Katz Activities of Daily Living (ADL) Scale,^9,23 the Blessed Dementia Rating Scale^10,16-19,24 and the Stockton Geriatric Rating Scale,^19,25 have been used. This is also the case outside the field of dementia.^26-31 None of these investigations, however, has brought together all of the important elements of mobility. An aggregation of these features that includes ADL, falls, walking, gait and motor system disorders would result in a powerful and instrumental definition of mobility. The combination of ADL measures and clinical variables would be unique in the study of mobility in dementia.

The high prevalence of dementia in older people, the rapid increase in the oldest sector of the population in industrialized countries and the lack of an effective treatment to reverse the disease underscore the need to identify factors related to favourable prognosis in dementia.⁹ The aim of this study, a two-year evaluation of 957 subjects with dementia, was to identify elements of mobility that predict survival in elderly patients with dementia.

Methods

Setting and Participants

The study involved subjects from the Canadian Study of Health and Aging - Part 1 (CSHA-1). The CSHA-1 was a survey conducted across Canada in 1991-1992 to estimate the prevalence of dementia and its subtypes; details of the study methods have been reported elsewhere.² In brief, the survey included a representative sample of people aged 65 and over from community and institutional settings in all 10 provinces. Thirty-six cities and their surrounding rural areas were selected by cluster sampling;³² roughly 60% of Canadians 65 and over lived in the sample areas.²

The community sample was obtained from the databases of the provincial health insurance plans, except in Ontario, where enumeration records were used. Of the 19,398 people on the community sample lists, 9,008 were eligible and agreed to participate.² The institutional sample comprised subjects in nursing homes, chronic care facilities and collective dwellings such as convents. It consisted of 1,817 subjects, of whom 1,255 were eligible and agreed to participate.²

For community participants, the survey had a two-phase design. In the first phase, individuals were interviewed in their homes after written consent had been obtained. The Older Americans Resources and Services ADL Scale was completed,³³ and a Modified Mini-Mental State (3MS) examination was performed.³⁴ Individuals scoring 77 or less on the 3MS (maximum score = 100) were identified as potentially cognitively impaired and were invited to proceed to the second phase of the study, which involved a clinical assessment.³² A random sample of those who scored higher than 77 was also invited to participate in the second phase in order to estimate the false-negative rate of the 3MS. Institutionalized subjects were evaluated in one phase, which included both the interview and the clinical assessment.

The clinical examination assessed the presence of cognitive impairment and provided a differential diagnosis of dementia.² The diagnostic criteria for dementia followed the third revision of the Diagnostic and Statistical Manual of Mental Disorders (DSM-III-R).³⁵ A diagnosis of Alzheimer's disease was based on the criteria of the National Institute of Neurological and Communicative Disorders and Stroke, and the Alzheimer's Disease and Related Disorders Association.³⁶ A draft of the 10th revision of the International Classification of Diseases was used to define subcategories of vascular and other dementias.³⁷

The examination was in four parts. Clinical team members were unaware of the 3MS score obtained at the screening interview. First, a nurse administered the 3MS examination; tested hearing, vision and vital signs; recorded height, weight and medication use; and obtained the subject's cognitive and family history from a relative, using section H of the Cambridge Mental Disorders of the Elderly Examination (CAMDEX).^2,38 Second, a neuropsychologist evaluated psychometric test results in conjunction with the results of the CAMDEX and the 3MS examination administered by the nurse. Third, a physician reviewed the information collected by the nurse and examined the patient, performing a mental status assessment as well as physical and neurologic examinations. Finally, subjects suspected of having dementia or delirium were sent for hematologic and biochemical tests.²

Final diagnostic categories included no cognitive impairment (NCI), cognitive impairment without dementia (CI) and four general types of dementia: AD (probable and possible Alzheimer's disease), VaD (vascular dementia), other dementia and unclassified dementia.³² Other dementias included those resulting from Huntington's chorea, Creutzfeldt-Jakob disease, Parkinson's disease, Pick's disease and head injury. The final diagnosis for each participant was reached in a consensus conference that integrated all available data on a particular individual and included the physician, study nurse, psychometrician and neuropsychologist who had examined the patient.

Consistency of diagnosis was assessed in two ways. First, there was a comparison of the diagnosis given with that obtained from a computer algorithm. Second, for 210 cases, a review of clinical chart data by a study physician from another centre was performed. The agreement between the computer classification and the final diagnosis was 97.6% for the distinction between dementia and non-dementia and 97.5% for the distinction between probable AD and other types of dementias.³² The kappa index of agreement between the study diagnosis and the review diagnosis by the second physician was 0.70 for the classification of NCI, CI, AD and other dementias.²

The present investigation followed a cohort of 957 subjects: 272 community subjects with a diagnosis of AD and 77 with a diagnosis of VaD; and, of institutionalized subjects, 477 with AD and 131 with VaD. Subjects with a diagnosis of "other" and "unclassified" dementia were excluded from the study.

In 1993-1994, two years after the CSHA-1, the 18 study centres conducted a field study, known as the Maintaining Contact Study (MCS), to re-contact CSHA-1 participants. Basic health status, including vital status, was assessed at that time. The CSHA-2 study then followed in 1996-1997, and re-evaluated the individuals involved in CSHA-1. The vital status of each individual in the CSHA-1 cohort was assessed using MCS data from two years after the clinical examination in conjunction with CSHA-2 data. Information on the date of death was obtained through next-of-kin interviews and confirmed with the Provincial Registrar of Births and Deaths. At this time, 267 subjects with AD and 81 subjects with VaD had died.

Prognostic Factors

Mobility

The main predictor of interest in this study was mobility. Items and measures were selected from the clinical examination phase of the CSHA-1 based on five broad categories: ADL, walking, falls, motor system disease and gait. These elements formed an operational definition of mobility.^{4,8-10,16-19,26,28,30}

The ADL category was represented by difficulties in dressing. This item was part of the CAMDEX section of the clinical assessment.³⁸ The CAMDEX interview was conducted by a study nurse with a relative, friend or caregiver who may or may not have been living with the subject. Nurses underwent a one-week training program in administration of the CAMDEX, which is both a valid and reliable measure.^2,38 Subjects were categorized as having, or not having, difficulties in dressing, which ranged from misaligned buttons to complete inability to dress themselves. Information on this variable was available for 926 of the 957 subjects. Walking and ability to get about were also assessed by the CAMDEX. Participants were classified as having, or not having, difficulties in getting about. Data on this variable were missing for 101 subjects.

Information on previous falls was obtained by the physicians through a clinical history involving the patient and/or family member. The physicians' clinical and physical assessments were standardized using training materials such as manuals and videotapes.² Information on history of falls was unavailable for 66 subjects. Motor system strength and gait were assessed by physical examination. These variables were categorized as normal, abnormal or could not be examined. Gait pattern and motor strength could not be examined in 34% and 13% of individuals respectively.

Covariates

Because other factors, for example age and comorbidity, are also associated with mobility, they are classically defined covariates of the mobility-survival association and were adjusted for in the analyses. Sociodemographic covariates were collected during the clinical assessment phase of the study. Data for age, sex and residence were complete. Information on marital status was missing for eight individuals, and educational data were available for 795 of the 957 participants.

Further information was gathered on disease-related covariates, including dementia type, duration and severity, and cognitive function (3MS). Subjects were categorized as having a diagnosis of AD or VaD, and dementia severity was graded by physicians as mild, moderate or severe; severity data were missing for 11 subjects. The duration of dementia was taken as the age of onset of first memory symptoms (CAMDEX data) subtracted from age at clinical assessment. This information was available for 861 of the 957 participants. The 3MS examination to evaluate cognitive function was conducted during the clinical assessment phase, and the score was summarized into four categories (0-29, 30-49, 50-69 and 70-100); scores were missing for 98 subjects.

Data on comorbid conditions were collected using physician clinical histories and the CAMDEX. Subjects were categorized as having a positive or negative history of heart attack, stroke, Parkinson's disease and respiratory difficulties (asthma, bronchitis). For 39 subjects, data were missing on heart attack; for 121, on respiratory problems; for 47, on stroke; and for 23, on Parkinson's disease.

Statistical Methods

Descriptive univariate analyses and comparisons among all prognostic variables were conducted using simple distributions, Pearson correlations and chi-squared tests.³⁹ Bivariate comparisons between individual predictors and outcome status were made using chi-squared tests.³⁹ The effect of prognostic factors on survival was then further evaluated by unconditional logistic regression modelling.⁴⁰ This technique provided the best fitting and most parsimonious models to describe the relation between the dichotomous survival outcome and the set of independent mobility and covariate variables.⁴⁰ Age, sex, education and type of dementia were substantively included in the models.

Each mobility variable was then modelled independently of the others. Possible prognostic covariates were entered successively. The four variables related to comorbid conditions were entered as a group. This modelling approach was used for the analysis of all demented subjects and for separate analyses of community and institution dwellers. Regression was performed using the indicator method to account for missing data:^41-43 for each variable with missing values a missing-value indicator was created,^41,42 which took the value 1 wherever the original variable was missing and 0 elsewhere.^41,42 Statistical analyses were conducted using the SAS statistical software package.⁴⁴

Mobility and several covariates were investigated as potential predictors of survival. Age was entered in the models as a continuous variable. Sex (male or female), residence (community or institution), education (< 8 years or 8+ years), dementia type (AD or VaD), difficulty dressing (yes or no), difficulty getting about (yes or no), history of falls (yes or no) and all comorbidity variables (yes or no) were evaluated as dichotomous. Finally, disease duration (< 2, 3-5, 6+ years), severity (mild, moderate, severe), 3MS score (0-29, 30-49, 50-69 and 70-100), gait (normal, abnormal, could not be examined) and motor strength (normal, abnormal, could not be examined) were evaluated as categorical variables.

Results

General Characteristics

Of the 957 study subjects, 749 had a diagnosis of AD and 208 of VaD. Mean age at clinical examination was 85.0 (standard deviation = 7.1) years. Table 2 lists the principal baseline characteristics of the sample, and Table 3 summarizes the baseline mobility and comorbidity data.

In further unadjusted analyses, several sociodemographic characteristics varied significantly between community and institution dwellers. Subjects living in institutions were older than subjects in the community (p = 0.005) and included a higher proportion of females (p < 0.001) and of severe cases of dementia (p < 0.001). However, no significant differences were observed in the proportion of more highly educated participants (p = 0.265) or in the proportion of AD versus VaD dementia (p = 0.852) as a function of residence.

The unadjusted proportion of individuals with difficulties in dressing and getting about differed significantly with respect to residence, as did the unadjusted proportion of subjects with past falls, gait problems and strength difficulties (all p < 0.001). A history of stroke and Parkinson's disease was found to be more common in institutionalized subjects (p = 0.003 and p = 0.050 respectively), whereas no significant differences in a history of heart attack (p = 0.444) or respiratory problems (p = 0.150) were observed between community and institution residence.

Correlation analysis of sociodemographic, mobility and comorbidity variables revealed a high Pearson correlation coefficient of 0.756 (p < 0.001) between disease severity and 3MS score. Because of a high occurrence of missing data for 3MS (10.2%) as compared with disease severity (1.1%), the 3MS variable was excluded from further analysis. Replacing severity with 3MS in the analyses yielded similar results.

Predictors of Survival

Unadjusted bivariate comparisons between individual predictors and outcome status revealed that survivors were significantly younger (p < 0.001) and were more likely to be female (p = 0.024), to live in the community (p < 0.001) and to have a milder dementia (p < 0.001) than those who died (Table 2). There was no significant difference between survivors and deceased in educational level (p = 0.704), duration of disease (p = 0.067) or type of dementia (p = 0.382). Evaluation of individual mobility variables indicated that survivors were less likely to have a history of falls (p < 0.001), difficulties in dressing (p < 0.001), difficulties in getting about (p < 0.001), gait problems (p < 0.001) and strength problems (p < 0.001) (Table3). There were no significant differences in comorbidity characteristics, including the proportion of participants with histories of stroke (p = 0.060), respiratory problems (p = 0.231) or Parkinson's disease (p = 0.082). However, a history of previous heart attack was found to be less likely in survivors after two years of follow-up (p = 0.047).

TABLE 5 Predictors of death in dementia after two years of follow-up: logistic regression models of community dwellers (n = 349)
Predictor	Model 1		Model 2		Model 3		Model 4		Model 5
	OR	95% CI	OR	95% CI	OR	95% CI	OR	95% CI	OR	95% CI
Age (years)	1.14	1.08-1.20c	1.14	1.08-1.21c	1.15	1.08-1.21c	1.14	1.07-1.20c	1.15	1.09-1.21c
Male vs female	2.54	1.40-4.61b	2.48	1.37-4.50b	2.66	1.46-4.84b	2.69	1.46-4.95b	2.55	1.40-4.64b
Education
Higher vs elementary	0.86	0.48-1.55	0.86	0.48-1.55	0.89	0.50-1.60	0.84	0.46-1.51	0.87	0.49-1.57
Dementia type
Vascular vs Alzheimer's	2.05	0.95-4.42	2.20	1.04-4.67a	2.18	1.03-4.60a	1.99	0.93-4.27	2.20	1.01-4.80a
Severity of dementia
Moderate vs mild	1.21	0.65-2.25	1.31	0.72-2.40	1.29	0.71-2.34	1.36	0.75-2.46	1.37	0.76-2.49
Severe vs mild	3.26	1.13-9.40a	3.83	1.43-10.2b	4.00	1.52-10.5b	3.89	1.47-10.3b	3.81	1.44-10.1b
Mobility
Difficulty dressing	1.67	0.85-3.28
Difficulty getting about			1.89	1.02-3.49a
History of falls					1.77	0.96-3.26
Gait (vs normal)
Abnormal							1.91	1.01-3.60a
Could not be examined							1.98	0.66-5.93
Motor strength (vs normal)
Abnormal									1.24	0.60-2.53
Could not be examined									2.24	0.41-12.2
Comorbidities
Heart attack	1.62	0.78-3.38	1.50	0.71-3.15	1.56	0.76-3.28	1.50	0.71-3.14	1.61	0.77-3.34
Respiratory problems	0.98	0.49-1.98	1.01	0.50-2.04	1.01	0.50-2.05	0.93	0.46-1.88	0.99	0.49-1.98
Stroke	0.73	0.35-1.55	0.70	0.33-1.46	0.71	0.34-1.48	0.68	0.33-1.43	0.70	0.33-1.47
Parkinson's disease	2.73	0.53-14.1	3.22	0.63-16.6	3.26	0.64-16.5	2.39	0.45-12.6	3.01	0.59-15.4
a 0.050 >= p >= 0.010 b 0.010 >= p >= 0.001 c p < 0.001

TABLE 6 Predictors of death in dementia after two years of follow-up: logistic regression models of institutionalized subjects (n = 608)
Predictor	Model 1		Model 2		Model 3		Model 4		Model 5
	OR	95% CI	OR	95% CI	OR	95% CI	OR	95% CI	OR	95% CI
Age (years)	1.03	1.01-1.06b	1.04	1.01-1.06b	1.03	1.01-1.06a	1.03	1.01-1.05a	1.03	1.01-1.06b
Male vs female	1.56	1.04-2.33a	1.58	1.06-2.36a	1.55	1.04-2.33a	1.59	1.06-2.37a	1.56	1.04-2.33a
Education
Higher vs elementary	1.18	0.80-1.73	1.17	0.79-1.72	1.20	0.82-1.77	1.22	0.83-1.80	1.22	0.83-1.80
Dementia type
Vascular vs Alzheimer's	0.82	0.50-1.36	0.81	0.49-1.34	0.82	0.49-1.36	0.81	0.49-1.34	0.74	0.44-1.23
Severity of dementia
Moderate vs mild	0.82	0.44-1.55	0.96	0.52-1.81	0.93	0.50-1.73	0.92	0.49-1.72	0.91	0.49-1.70
Severe vs mild	0.82	0.43-1.57	1.04	0.56-1.94	1.09	0.59-2.02	0.98	0.53-1.83	1.02	0.55-1.90
Mobility
Difficulty dressing	2.12	1.28-3.51b
Difficulty getting about			1.59	1.02-2.50a
History of falls					1.38	0.97-1.98
Gait (vs normal)
Abnormal							1.29	0.74-2.23
Could not be examined							1.75	1.00-3.05a
Motor strength (vs normal)
Abnormal									1.65	1.09-2.49a
Could not be examined									1.38	0.84-2.27
Comorbidities
Heart attack	1.54	0.93-2.53	1.64	0.99-2.71	1.54	0.94-2.54	1.57	0.95-2.59	1.63	0.99-2.70
Respiratory problems	1.47	0.88-2.46	1.25	0.75-2.07	1.39	0.84-2.31	1.33	0.81-2.21	1.30	0.79-2.16
Stroke	1.25	0.81-1.93	1.24	0.81-1.92	1.23	0.80-1.91	1.21	0.78-1.87	1.18	0.76-1.82
Parkinson's disease	1.45	0.67-3.14	1.36	0.63-2.94	1.39	0.64-2.99	1.40	0.65-3.01	1.44	0.67-3.10
a 0.050 >= p >= 0.010 b 0.010 >= p >= 0.001 c p < 0.001