Caregiver Reports of Sleep Problems in Non-Hispanic White, Hispanic, and African American Patients with Alzheimer Dementia

Abstract

Study Objectives:

Sleep problems are common in persons with dementing illnesses and among the most stressful patient behaviors for caregivers. Although studies have shown differences in sleep across ethnic groups, little information is available on ethnic differences among persons with dementia. The purpose of this study was to investigate possible ethnic differences in sleep problems among patients with Alzheimer dementia.

Method:

Caregiver reports of 5 sleep- or circadian rhythm-related behavioral problems (behavior disturbance worse in the evening, difficulties falling asleep, frequent awakenings, early awakenings, and excessive daytime sleep) were evaluated in 395 patients who had received a diagnosis of Alzheimer disease after diagnostic evaluation. The average cognitive score of the groups suggested that they could be characterized as having moderately severe impairment. The frequency of sleep problems was then evaluated across subgroups defined by self-reported ethnicity (African American, Hispanic, and non-Hispanic white). As patient and caregiver characteristics may affect caregivers' reports of patients' behaviors, mixed effects regression models were used to adjust for patient and caregiver variables that might affect caregiver reports.

Results:

Analyses revealed ethnic differences in sleep or circadian rhythm disturbances. African American and Hispanic patients were reported to have more severe sleep disturbances than non-Hispanic whites. After correction for patient and caregiver variables that might have affected caregiver reports, differences between African Americans and others remained.

Conclusions:

Sleep problems in patients with dementing illnesses are reported by caregivers with differing frequencies across groups of African Americans, Hispanics, and non-Hispanic whites. Clinicians should be aware of these differences in assessing sleep disturbance in patients with dementia as well as the potential effects of patient and caregiver variables on reports of these problems.

Citation:

Ownby RL; Saeed M; Wohlgemuth W; Capasso R; Acevedo A; Peruyera G; Sevush S. Caregiver reports of sleep problems in non-Hispanic white, Hispanic, and African American patients with Alzheimer dementia. J Clin Sleep Med 2010;6(3):281-289.

As the average age of the population increases and as age is one of the most important risk factors for dementia, the prevalence of dementing illnesses is also increasing. The cost of care for patients with dementia is estimated to be in the billions of dollars¹ and is expected to increase fourfold over the next 40 years.¹ Dementia caregiving costs arise from multiple sources but are based in part on the need to provide protective care for cognitively impaired patients. A large portion of the costs associated with dementing illnesses arises from patient behavior problems that include agitation, depression, and sleep disturbance.^1–3

Brief Summary

Current Knowledge/Study Rationale: Sleep and circadian rhythms are commonly affected by Alzheimer's disease, resulting in significant impairment of patients' functioning and increased caregiver stress. Little is known, however, about racial or ethnic differences in sleep in this patient population.

Study Impact: Caregivers of Hispanic and African-American patients reported that patients had more severe sleep disturbances than for non-Hispanic whites. Clinicians should be aware of the possible presence of racial and ethnic differences in sleep disturbances in working with patients with Alzheimer's disease.

Sleep disturbances are common in these patients and may be related to disturbances in circadian rhythms seen in these patients. Especially in patients with Alzheimer disease (AD), patients display worsening confusion in the evening, frequent nocturnal awakenings, and increased daytime sleepiness.⁴ Nocturnal awakenings are associated with increased risk for confusion and agitation and also increase the risk for wandering. Sleep problems are among the most troublesome of problems for dementia caregivers and are related to the likelihood that a patient will be institutionalized.⁵ Institutionalization of these patients is in turn associated with reduced quality of life and increased cost of care. Sleep disturbance in patients with AD is thus associated with increased patient depression and poorer functional status.⁶ Poorer functional status is itself associated with increased caregiving costs and increased caregiver burden.^7,8 Sleep problems in patients with AD may thus be important because of their effect on patients' mood and functional status⁹ and the related effects of these factors on caregivers and the economic costs of providing care for affected individuals.

In addition to affecting patients' mood and functional status, sleep disturbances may have an impact on their safety. It can be speculated that nighttime awakenings and related unsupervised wandering may put them at risk for inadvertent injury and increase caregiver burden through the demands they place on caregiver time and sleep disruption. Nighttime wandering and awakening caregivers at night are two sleep related problems that may increase caregiver burden. Sleep disturbances are also associated with increased healthcare costs. One study, for example, estimated that sleep related institutionalization of dementia patients is associated with a large proportion (more than $10 billion) of all costs associated with insomnia.¹⁰ Another study showed that sleep problems were an important factors in caregivers' decisions to institutionalize patients.¹¹ A better understanding of sleep problems might thus help to improve both patient and caregiver health, alleviate a significant source of caregiver burden, and decrease health care costs by helping caregivers maintain patients in the community rather than placing them in institutions.¹²

As noted, sleep disturbances in AD include increased frequency and length of nocturnal awakenings and increased frequency of daytime napping.⁴ Basic circadian rhythms are disrupted in patients with AD and this disruption is associated with the severity of patients' sundowning (an increase in confusion and agitation in the early evening).¹³ Sleep disturbances in AD are consistent with the underlying neuropathology of the disease in that the regions of the brain most severely affected by AD (subcortical areas including aminergic and serotonergic nuclei) are also implicated in basic sleep mechanisms.^14,15 Sleep disturbances are associated with poorer cognitive and functional status in patients with AD.^16,17 Sleep disturbance is thus an important component of the behavioral disturbances associated with AD, and are related to increased caregiver burden.⁵ Available data suggest that the prevalence of sleep problems in nondemented individuals may vary across ethnic groups including non-Hispanic whites, Hispanics,¹⁸ and African Americans.¹⁹ Ethnic or racial differences have been observed in the occurrence of sleep disordered breathing, duration of sleep,^19–21 periodic limb movements,²² sleepiness,²³ and in polysomnographic indexes that have included time spent in specific sleep stages and, in persons with depression, rapid eye movement sleep (REM).²⁴ Differences in sleep may mediate some disparities in health outcomes for persons in minority groups. A well-established risk factor for developing AD, the ApoE4 genotype, is also related to the risk for sleep related breathing disorders²⁵ and the course of sleep problems in patients with dementia.²⁶ Persons with this genotype may thus be at risk for developing cognitive impairments via several mechanisms.

The psychological and medical effects of dementia caregiving (increased risk of physical illnesses and increased risk of death) have been clearly established.²⁷ Caregivers are more likely than non-caregivers to suffer from depression and require treatment for it,²⁸ while caregiving burden is related to increased frequency of health problems,²⁹ as well as increased risk for mortality compared to non-caregivers.³⁰ Dementia caregivers rate sleep problems as one of the most troublesome aspects of caring for a person with the disease.⁵ In spite of the importance of sleep disturbance for patients with dementia and their caregivers, little information is available on the prevalence of sleep problems across ethnic groups. Although studies have often included African American or Hispanic patients as well as non-Hispanic whites, most studies have not reported data on sleep behaviors separately for racial or ethnic groups.

Important issues to consider in evaluating the occurrence of sleep problems among patients with dementing illnesses concern measurement. Perhaps the simplest approach is to ask patients' spouses or other caregiver about the occurrence of behaviors suggesting sleep disturbance. This strategy is complicated by the fact that caregivers' own characteristics, such as age or mood, may affect their reports of problems.^31,32 An additional issue in this context is the likelihood that patient characteristics may affect caregivers' reports of sleep problems. Worse patient functional status or patient depression may increase the likelihood that caregivers will report problematic behaviors. Any evaluation of the occurrence of sleep problems in patients with dementia should therefore take these issues into account.

The purpose of this study was thus to evaluate the frequency and severity of caregiver-reported sleep problems in African American, Hispanic, and non-Hispanic white patients with Alzheimer disease.

METHODS

The clinic at which data for this study were collected provides evaluation and treatment services for individuals with memory disorders and their families. Services include diagnostic evaluations, ongoing treatment, and referral to community resources. Patients are referred by community physicians, community service agencies, and by themselves or family members. The clinic was established in 1978 and has operated continuously since then. Patients are drawn from the Miami-Dade county area as well as from throughout South Florida. Services are provided both in Spanish and English, and a large proportion of clinic patients are primarily Spanish speaking.

As part of routine initial and follow-up evaluation visits at 6-month intervals, patients complete an evaluation that includes a battery of cognitive, functional, and mood measures, as well as neuropsychiatric and neurological evaluations. Most patients come for evaluation accompanied by a caregiver, who most frequently is a spouse or adult child. Caregivers complete several questionnaires providing information about the patients' behavior and their own mood. All patients for whom informant data are available arrived at the clinic with an informant who could provide information, or the clinic team was able to contact an appropriate informant by telephone. Instances in which the patient arrived without an informant were extremely rare and would have resulted in missing data; these cases were not included in analyses. Caregiver assessments were done at the same visit as the patient assessment. Descriptive data presented in Table 1 reflect initial evaluations. Patients were seen at approximately 6-month intervals. Data for regression models include all available data for each patient in order to increase the reliability and validity of analyses.

Table 1.

Descriptive statistics and group differences by ethnicity^a

	Non-Hispanic White	Hispanic	African American	Total Sample	df	F	p
N	104	252	39	395
Patient age	77.0 (7.9)	76.1 (8.7)	78.6 (6.4)	76.1 (8.3 )	2, 392	2.85	0.059
Patient education (Years)	13.4 (3.5)	9.3 (4.7 )	10.0 (4.)7	10.4 (4.7)	2, 392	32.22	< 0.001
MMSE	15.1 (7.3)	12.9 (5.8)	11.5 (6.0)	13.3 (6.3)	2, 390	6.68	0.0014
Blessed	5.2 (3.7)	6.6 (3.8)	6.9 (3.7 )	6.3 (3.8)	2, 391	5.66	0.004
Patient CSDD	4.8 (5.2)	5.3 (5.9)	3.4 (3.3)	5.0 (5.5)	2, 364	2.13	0.12
Caregiver CESD	9.7 (8.0)	11.6 (9.4)	10.1 (6.9)	11.0 (8.8)	2, 382	1.88	0.15
BMI	23.9 (2.3)	25.3 (3.8)	25.4 (2.7)	25.0 (3.4)	2, 212	3.42	0.03

^aValues in the table are means and standard deviations (in parentheses). Group differences in potential confounding variables were evaluated with one-way ANOVAs. Degrees of freedom vary because of missing data for some variables. MMSE, Mini-mental State Exam; Blessed, Blessed Dementia Rating Scale (note that higher values for this scale indicate worse functioning); CSDD, Cornell Scale for Depression in Dementia completed with patients; CESD, Center for Epidemiological Studies Depression Scale completed with caregivers; BMI, body mass index

After the initial evaluation, the team, comprising a neuropsychiatrist and trained gerontological assessors, meets to review assessment results and establish a consensus diagnosis using DSM criteria (DSM-III for patients seen in the 1980s; DSM-IV after it became available).^33,34 Diagnoses are re-evaluated at each follow-up visit and updated as indicated by assessment results. The majority of patients seen in the clinic receive a diagnosis of probable AD, with a substantial minority given the diagnosis of vascular dementia. Although other causes of dementia (e.g., Lewy body or frontotemporal dementias) are seen in the clinic, because of the small number of patients with each diagnosis and their distinct neuropathologies, only patients with probable AD were included in this study.

The clinic is a teaching site for a university-based psychiatry residency and geriatric psychiatry fellowship program. During the period of data collection, patient psychiatric assessments were completed by bilingual geriatric psychiatry fellows when patients were evaluated in Spanish. Paid evaluators trained and supervised by the clinic director completed elements of the cognitive evaluation (such as the Mini-Mental State Exam).³⁵

For the purpose of this study, we extracted information from the clinic database on patients' cognitive status on the Mini-mental State Exam (MMSE)³⁵ and functional status on the Blessed Dementia Rating Scale (Blessed).³⁶ The MMSE is a brief cognitive assessment measure widely used in geriatric assessment and includes items that tap memory (remembering 3 words) attention (subtracting 7 repeatedly from a base number) and visuospatial skills (copying a geometric design). Higher scores on the MMSE indicate better cognitive function. Scores on this measure range from 0 to 30; 23 is a commonly used cut-off score for the presence of cognitive impairment. The Blessed has also been used extensively in geriatric assessment and allows an assessment of the extent to which patients can perform key self-care skills independently. It includes items evaluating, for example, whether the patient can manage small amounts of money and find his or her way around the neighborhood, as well as items evaluating changes in core self-care skills such as eating and dressing. Higher scores on this measure indicate worse functional status.

Because of the key role of depression in sleep disturbance, patients' scores on the Cornell Scale for Depression in Dementia (CSDD)³⁷ were also recorded. The CSDD is a depression rating scale specifically developed to evaluate depressive symptoms that are seen in patients with dementing illnesses. These include items evaluating mood, anxiety, sleep and appetite disturbance, suicidality, and psychosis. It is completed based on patient self report and clinical evaluation. Data on patients' demographic characteristics included their self-reported ethnicity, age, height and weight, and gender.

Sleep variables were drawn from a standard questionnaire completed by all caregivers on patients' behavior problems. This questionnaire was developed for use in the clinic to allow a broad assessment of both common and uncommon behavioral problems that may be of clinical concern in patients with dementia. Five questions on this measure assessed sleep or circadian-rhythm associated problems, including items asking about whether behavior problems are worse in the evening (i.e., sundowning), whether the patient has difficulty getting to sleep, wakes frequently during the night, wakes early, and sleeps excessively during the day. Caregivers rated each of the behaviors on a scale from 0 (“never/not at all”) to 3 (“always/severe”). As caregiver reports of patient behavior can be influenced by caregiver characteristics,^31,32 caregiver mood was assessed using the Center for Epidemiologic Studies Depression (CES-D) scale.³⁸ Caregiver gender was also extracted from the database. Retrospective data analyses were completed under a protocol approved by the University of Miami Office of Human Subjects Protection.

Data Analyses

In order to evaluate the frequency and severity of caregivers' reports of sleep problems in patients across ethnic groups, as preliminary steps between-group differences on continuous variables representing patient and caregiver characteristics were assessed with one-way analyses of variance (ANOVA) and on categorical variables with χ² analysis. Mean ratings on each of the 5 problems across ethnic groups were also evaluated with one-way ANOVA. These analyses allowed us to present mean unadjusted values on the variables across groups prior to adjustment so that the reader can evaluate the groups' baseline characteristics. These models, however, did not allow for a statistical adjustment based on variables that have been shown to affect caregivers' reports of patient behavior, such as caregiver depression; the models also did not take into account possible differences between patients with AD and vascular dementia.

To address these issues and to maximize our power to detect group differences by using data collected on the same patients over several assessments, multivariate mixed effects models were utilized. We were able to take advantage of data from multiple visits by developing mixed effects regression models that allowed an assessment of group differences in sleep problems while accounting for repeated measures on each patient. This approach allowed us to correct for factors such as patient and caregiver characteristics while using all data available from patient visits over time. Models predicting severity of each problem while correcting for relevant patient and caregiver variables were thus created using the routine PROC MIXED in the SAS statistical software (Cary, NC).³⁹ This strategy allowed an assessment of the effects of patient and caregiver variables over time while accounting for baseline differences and individual variability over time⁴⁰ using random intercepts and random slopes for time. Data were available for 395 patients, with an average of 2.5 visits per patient over 1-2 years. Variables included in these models were patient cognitive and functional status, patient age, gender, and depression, and caregiver gender and depression, all of which have either been previously shown to affect caregiver reports of patients' behavior or which differed across groups in preliminary one-way ANOVAs. Consistent with standard practice in the multiple regression analysis, ethnicity was dummy coded as 0, 1, or 2. Whites were included in analyses although a term for whites is not included in the tables reporting the analyses.⁴¹

Although no measure that directly assessed sleep disordered breathing was included in the clinic evaluation, body mass index (BMI) is an important risk factor for its occurrence^21,42 and was calculated for a subset of clinic patients for whom height and weight had been recorded. BMI was defined as kilograms per square meter of body area made operational as patients' weight in pounds times 703 divided by their height in inches squared.

In preliminary analyses, the effect of BMI, along with gender and age, were studied since they are among the most important risk factors for sleep disordered breathing.⁴³ Since older African Americans may be at increased risk for sleep disordered breathing,⁴⁴ it was particularly important to evaluate this issue. Because inclusion of BMI substantially reduced the sample available for analysis to those for whom heights and weights had been recorded, and its inclusion in regression models did not change conclusions (all probability values for BMI coefficients in regression models for the 5 sleep behaviors were > 0.40), results presented here do not include BMI.

Because of the importance of caregivers' reports of patients' behavior, we evaluated the relation of caregiver and patient gender to caregiver reports of behavioral disturbances in our analyses.

Although caregiver age was not recorded in the clinic, it was possible to infer relative age of the caregiver compared to the patient based on the relationship of caregiver and patient. Caregivers listed as spouses were assumed to be of an age similar to the patient, while caregivers who were children or grandchildren were assumed to be younger. This variable was used in multiple regression analyses to allow for the potential effect of caregiver age in reports of patient behavior.

It was possible that there might be an effect of the match of caregiver and patient genders. This possibility was evaluated in regression models reported below as an interaction between caregiver and patient gender. The possibility that ethnicity and patient and caregiver gender were related was also evaluated by including the 3-way interaction term in regression models. As these interaction effects were not statistically significant in any of the regression models, they are not reported here.

RESULTS

Descriptive statistics for the sample and subgroups based on ethnicity at the initial evaluation are provided in Table 1. Baseline differences on these variables were assessed with one-way analyses of variance (ANOVA) in order to evaluate the equivalence of groups. Group differences were found in years of education, cognitive function (MMSE), functional status (Blessed), and body mass index (BMI).

Given the importance of caregiver report in understanding the results of this study, the relation of caregiver informants to the patient evaluated was tabulated. This tabulation is presented in Table 2. Most informants were either spouses or the sons or daughters of patients, suggesting that in most cases the caregiver might have been reasonably knowledgeable about the patient's status.

Table 2A.

Informant relationships to patient

Relationship	Frequency	Percent
Aide	2	0.50
Brother	3	0.76
Daughter	140	35.44
Friend	5	1.27
Grandchild	7	1.78
Husband	95	24.05
Landlady	1	0.25
Niece	7	1.78
Sister	11	2.78
Son	45	11.17
Wife	79	20.00

Table 2B.

Relation of patient and caregiver genders

	Caregiver Gender
Patient Gender	Men	Women	Total
Men	19	82	101
Women	267	23	290
Total	286	105	391

One-way ANOVAs of each behavioral problem across groups are presented in Table 3. Between-group effects were found for each of the behavior problems, with the exception of excessive daytime sleepiness. No between-group post hoc comparisons for behavior worse in the evening reached the level of statistical significance, in spite of the overall effect. African Americans were reported to have more severe or frequent disturbances of sleep for several variables (difficulty falling asleep and early morning awakenings) compared with non-Hispanic whites. Ratings for Hispanics were higher for early morning awakenings when compared to non-Hispanic whites.

Table 3.

Mean ratings of sleep variables across ethnic groups^a

	Non-Hispanic White	Hispanic	African American	df	F	p
Total n = 395	104	252	39
Behavior worse in the evening	1.28 (1.04)a	1.54 (1.12)	1.74 (1.12)	2, 392	3.23	0.041
Difficulty falling asleep	0.81 (1.01)	1.02 (1.12)b	1.31 (1.26)b	2, 392	2.97	0.053
Frequent awakenings at night	1.25 (0.98)	1.30 (1.12)	1.69 (1.13)	2, 388	2.56	0.078
Early morning Awakenings	0.94 (1.03)	1.10 (1.06)c	1.61 (1.20)c	2, 388	5.37	0.005
Excessive daytime sleeping	1.00 (1.07)e	1.00 (1.08)	0.92 (0.94)	2, 387	0.10	0.900

^aValues represent means and standard deviations (in parentheses). No between-group differences were statistically significant after correction for multiple comparisons. The p value for the difference between non-Hispanic whites and African Americans was 0.07 and for the difference between non-Hispanic whites and Hispanics was 0.10 after Tukey correction for multiple corrections.

^bRating for African Americans was significantly greater than that for non-Hispanic whites (p = 0.0487 ) after Tukey correction.

^cRatings for Hispanics and African Americans were significantly greater than that for non-Hispanic whites (p = 0.02 and p = 0.003, respectively). The difference between Hispanics and African Americans was not significant (p = 0.40).

Follow-up evaluations were available for patients approximately at 6-month intervals, although data from fewer patients were available at successive visits. We used data from 395 patients at time 1; 221 patients at evaluation 2; 148 at evaluation 3; 115 patients at evaluation 4; 87 patients at evaluation 5; and 64 patients at evaluation 6 in regression models that evaluated the relation of ethnicity to reports of sleep problems over time. Mixed effect models were constructed for each sleep problem, correcting for variables that might have affected caregivers' reports of sleep problems. These included patient cognitive and functional status, gender, education, age, and depression, as well as caregiver depression and gender.

Models for each sleep or circadian rhythm related behavior are presented in Tables 4–8. Ethnic differences were no longer apparent after correction for patient and caregiver variables for the variable behavior worse in the evening (Table 4). Patient functional status (Blessed) and caregiver depression (CESD) were significantly related to caregivers' reports of this problem. Women caregivers were more likely to report this behavior, and caregivers reported this behavior more frequently for female than male patients.

Table 4.

Regression model for worsening behavior in the evening^a

Effect	Coefficient	Standard Error	df	t	p
Intercept	−0.2533	0.5291	403	−0.48	0.632
Age	0.0097	0.0055	261	1.77	0.079
MMSE	−0.0110	0.0078	261	−1.41	0.160
Blessed	0.0618	0.0128	261	4.84	< 0.00001
Education	−0.0044	0.0083	261	−0.52	0.602
CSDD	0.0087	0.0065	261	1.33	0.183
CESD	0.0254	0.0043	261	5.96	< 0.0001
CG Gender	0.4589	0.1352	261	3.39	0.0008
CG Age	−0.1645	0.0961	261	−1.71	0.088
Patient gender	0.3375	0.1324	261	2.55	0.011
African Americanb	0.0374	0.1496	261	0.25	0.803
Hispanicb	0.1313	0.09795	261	1.34	0.181

^aMMSE, Mini-mental State Exam; Blessed, Blessed Dementia Rating Scale; CSDD, Cornell Scale for Depression in Dementia for patients; CESD, Center for Epidemiological Studies Depression Scale for caregivers; CG Gender, caregivers' gender coded so that a positive coefficient indicates that women caregivers were more likely to report the behavior; CG age, informant age similar or younger than the patient, coded so that a negative coefficient indicates that younger caregivers were more likely to report the behavior; Patient gender, patient gender coded so that a positive coefficient indicates that women were more likely to display the behavior.

^bTests associated with these factors reflect difference from base category which was non-Hispanic whites.

Table 8.

Regression model for excessive daytime sleepiness^a

Effect	Coefficient	Standard Error	df	t	p
Intercept	−0.761	0.5750	402	−1.32	0.186
Age	0.0118	0.0060	255	1.94	0.053
MMSE	0.0024	0.0081	255	0.29	0.770
Blessed	0.0603	0.0130	255	4.63	<0.0001
Education	0.0026	0.0088	255	0.30	0.767
CSDD	0.0253	0.0068	255	3.75	0.0002
CESD	0.0113	0.0044	255	2.60	0.010
CG Gender	0.1512	0.1451	255	1.04	0.299
CG Age	0.0461	0.1043	255	0.44	0.659
Patient Gender	0.4251	0.1427	255	2.98	0.003b
African Americanc	−0.0344	0.1653	255	−0.21	0.836d
Hispanicc	−0.1818	0.1077	255	−1.69	0.093

^aMMSE, Mini-mental State Exam; Blessed, Blessed Dementia Rating Scale; CSDD, Cornell Scale for Depression in Dementia for patients; CESD, Center for Epidemiological Studies Depression Scale for caregivers; CG Gender, caregivers' gender coded so that a positive coefficient indicates that women caregivers were more likely to report the behavior; CG age, informant age similar to or younger than the patient, coded so that a negative coefficient indicates that younger caregivers were more likely to report the behavior; Patient gender, patient gender coded so that a positive coefficient indicates that women were more likely to display the behavior.

^bTests of the difference between model-predicted average ratings for this variable showed that men were rated higher on this variable than were women (mean rating 1.26 vs. 0.84; (t(df = 255) = 2.98, p = 0.003).

^cTest associated with these factors reflect differences from base category which was non-Hispanic whites.

^dNo comparisons based on ethnicity were statistically significant.

Analyses for reports of problems falling asleep are presented in Table 5. As before, caregiver reports of this behavior were associated with worse functional status and higher levels of patient and caregiver depression (although the probability associated for the effect of patient depression was exactly 0.05). Younger informants (in general, sons or daughters) were more likely to report the behavior, and African American patients were more likely to be rated as having this problem.

Table 5.

Regression model for difficulty falling asleep^a

Effect	Coefficient	Standard Error	df	t	p
Intercept	−0.6901	0.5553	402	−1.24	0.215
Age	0.0059	0.0058	258	1.02	0.309
MMSE	0.0146	0.0079	258	1.86	0.064
Blessed	0.0766	0.0129	258	5.96	< 0.0001
Education	−0.0037	0.0086	258	−0.43	0.671
CSDD	0.0130	0.0066	258	1.97	0.050
CESD	0.0273	0.0043	258	6.36	< 0.0001
CG Gender	0.1517	0.1407	258	1.08	0.282
CG Age	−0.2022	0.1007	258	−2.01	0.046b
Patient Gender	0.0618	0.1382	258	0.45	0.655
African Americanc	0.4743	0.1599	258	2.97	0.003d
Hispanicc	0.1223	0.1039	258	1.18	0.240

^aMMSE, Mini-mental State Exam; Blessed, Blessed Dementia Rating Scale; CSDD, Cornell Scale for Depression in Dementia for patients; CESD, Center for Epidemiological Studies Depression Scale for caregivers; CG Gender, caregivers' gender coded so that a positive coefficient indicates that women caregivers were more likely to report the behavior; CG age, informant age similar to or younger than the patient, coded so that a negative coefficient indicates that younger caregivers were more likely to report the behavior; Patient gender, patient gender coded so that a positive coefficient indicates that women were more likely to display the behavior.

^bThis effect suggested that younger caregivers (generally the son or daughter of the patient) were less likely to rate this item highly).

^cTest associated with these factors reflect difference from base category which was non-Hispanic whites.

^dTests of differences between model-predicted average ratings showed that African Americans were rated significantly higher on this variable than were non-Hispanic whites (t(df = 258) = 2.97, p = 0.003). The difference between African Americans and Hispanics was also significant (t(df = 258) = 2.40, p = 0.017), but the difference between non-Hispanic whites and Hispanics was not significant (t(df = 258) = 1.18, p = 0.24).

Patient functional status and caregiver depression were also related to reports of frequent nighttime awakenings (Table 6). Test of model-corrected group means showed that African Americans were rated significantly higher on this item than were Hispanics, although the regression coefficient in the model for this effect was not statistically significant.

Table 6.

Regression model for frequent nighttime awakenings^a

Effect	Coefficient	Standard Error	df	t	p
Intercept	0.5798	0.5659	399	1.02	0.306
Age	0.0011	0.0060	256	0.18	0.858
MMSE	0.0013	0.0077	256	0.17	0.865
Blessed	0.0663	0.0125	256	5.31	< 0.0001
Education	0.5798	0.5659	399	1.02	0.306
CSDD	0.0115	0.0065	256	1.78	0.076
CESD	0.0276	0.0042	256	6.58	< 0.0001
CG gender	0.0441	0.1426	256	0.31	0.757
CG age	−0.1707	0.1022	256	−1.67	0.096
Patient gender	0.1477	0.1413	256	1.04	0.297
African Americanb	0.2799	0.1643	256	1.70	0.090c
Hispanicb	−0.1267	0.1067	256	−1.19	0.236

^aMMSE, Mini-mental State Exam; Blessed, Blessed Dementia Rating Scale; CSDD, Cornell Scale for Depression in Dementia for patients; CESD, Center for Epidemiological Studies Depression Scale for caregivers; CG Gender, caregivers' gender coded so that a positive coefficient indicates that women caregivers were more likely to report the behavior; CG age, informant age similar to or younger than the patient, coded so that a negative coefficient indicates that younger caregivers were more likely to report the behavior; Patient gender, patient gender coded so that a positive coefficient indicates that women were more likely to display the behavior.

^bTest associated with these factors reflect difference from base category which was non-Hispanic whites.

^cTests of differences between model-predicted average ratings showed that African Americans were rated significantly higher on this variable than were Hispanics (t(df = 256) = 2.70, p = 0.007), even though the overall effect was not statistically significant.

African Americans were more likely than Hispanics or non-Hispanic whites to be reported to display early morning awakenings (Table 7). As with other models, patient functional status and patient and caregiver depression were significantly related to reports of this behavior. Women patients were more likely to be reported to display excessive daytime sleepiness (Table 8). Patient functional status, patient depression, and caregiver depression were also related to reports of this problem. No ethnic differences were observed for this problem.

Table 7.

Regression model for early morning awakening^a

Effect	Coefficient	Standard Error	df	t	p
Intercept	0.4499	0.5373	399	0.84	0.403
Age	0.0045	0.0056	256	0.80	0.427
MMSE	−0.0018	0.0079	256	−0.23	0.818
Blessed	0.0482	0.0130	256	3.72	0.0002
Education	−0.0168	0.0085	256	−1.98	0.048
CSDD	0.0088	0.0067	256	1.33	0.186
CESD	0.0221	0.0043	256	5.09	< 0.0001
CG Gender	−0.0761	0.1369	256	−0.56	0.579
CG Age	−0.1443	0.0975	256	−1.48	0.140
Patient Gender	−0.2309	0.1343	256	−1.72	0.087
African Americanb	0.3504	0.1524	256	2.30	0.022c
Hispanicb	0.0743	0.0994	256	0.75	0.455

^aMMSE, Mini-mental State Exam; Blessed, Blessed Dementia Rating Scale; CSDD, Cornell Scale for Depression in Dementia for patients; CESD, Center for Epidemiological Studies Depression Scale for caregivers; CG Gender, caregivers' gender coded so that a positive coefficient indicates that women caregivers were more likely to report the behavior; CG age, informant age similar to or younger than the patient, coded so that a negative coefficient indicates that younger caregivers were more likely to report the behavior; Patient gender, patient gender coded so that a positive coefficient indicates that women were more likely to display the behavior.

^bTest associated with these factors reflect difference from base category which was non-Hispanic whites.

^cTests of differences between model-predicted average ratings showed that African Americans were rated higher on this variable than were non-Hispanic whites (t(df = 256) = 2.30, p = 0.022). The difference between Hispanics and African Americans approached statistical significance (t(df = 256) = 1.97, p = 0.0502).

DISCUSSION

This study investigated the presence of ethnic differences in caregiver reports of the frequency and severity of sleep and circadian rhythm disturbances in patients with dementing illnesses. Results of uncorrected one-way ANOVAs suggested that several sleep and circadian rhythm problems occurred with greater frequency or intensity in African American and Hispanic patients. The effect of ethnicity continued to be present only for problems among African Americans after correction for patient and caregiver variables. The findings confirm previous observations of the effects of patient and caregiver depression on caregiver reports of sleep problems in patients with dementia³² and others' observation of the relation between sleep disturbances and functional status.¹⁷

The association of caregiver reports of sleep disturbances with patient and caregiver depression suggests that two processes may interact in complex ways to result in caregiver reports. Caregiver depression is known to affect reports of behavior problems in patients, perhaps reflecting increased burden associated with worse patient functional status and the patients' own level of depression. Caregivers experiencing high levels of caregiving burden may be predisposed to report more frequent or severe behavioral disturbances because of the amplifying effect of poor mood on perceptions of problematic patient behavior. A second process might reflect the effect of patients' depression on sleep, which might be expected to be associated with difficulties falling asleep, frequent awakenings, and early morning awakenings. Reports of excessive daytime sleepiness might then reflect both depressed caregivers' propensity to report behavioral disturbance and patients' actual daytime sleepiness resulting from disrupted nocturnal sleep.

Even after correction for a number of variables that might have affected caregivers' reports of sleep disturbances, African Americans were rated as displaying more frequent or severe disturbance on several sleep related behaviors compared with either Hispanics or non-Hispanic whites. Other studies have documented the presence of differences in sleep among African Americans and Hispanics, although not in patients with dementia. Sleep related breathing disorders may be more common in African Americans than in other groups,⁴⁵ and undetected or subclinical sleep related breathing disorders might account for some sleep disruption in dementia patients. Sleep disordered breathing may also be related to poorer cognitive functioning in patients with dementia.⁴⁶ Unfortunately, our data do not allow us to adequately assess the possibility that sleep disordered breathing was an important factor in cognitive decline in our sample.

Another well-established factor related to sleep disturbance is depression.⁴⁷ This observation is confirmed in the mixed effects models reported here, as patient depression was consistently related to reports of sleep disturbances. Even after correcting for depression, however, the effect of African American ethnicity remained significant, suggesting that the ethnic differences observed were not solely due to depression. As already noted, caregiver depression may be related to their reports of problematic patient behavior.³¹ Because caregiver depression might be related to gender, we corrected for both gender and depression in regression analyses. Here again, although the effect of caregiver gender was significant in several analyses, the effect of African American ethnicity remained.

Several patient characteristics were consistently associated with sleep. The consistent association of patient depression with sleep disturbance is unsurprising but illustrates the potential importance of assessing and treating patient depression when addressing sleep disturbance. Although one study has shown that sleep problems in older persons do not always improve with treatment for depression and may be related to continuing depression,⁴⁸ commonly used pharmacologic treatments for depression (e.g., serotonin reuptake inhibitors) may improve ratings of depression and other behavior problems in patients with dementia.⁴⁹ Other behavioral,^50,51 biologic (i.e., bright light),⁵² and pharmacologic⁵³ interventions may also be useful in improving sleep in patients with dementia.

Reasons for the consistent association of functional status with sleep problems are unclear. Other research has shown that sleepiness is associated with both worse cognitive¹⁶ and functional status in patients with AD.¹⁷ Sleep disturbances may thus result in impairment in patients' functional status by decreasing their motivation to perform self-care activities or by increasing their confusion. Since worse functional status is likely to be linked to increased caregiver burden, this association may also reflect the effects of functional status on caregivers' reports of sleep disturbance via increases in burden. Caregivers may be inclined to encourage patients with poor functional status to spend more time in bed or napping, resulting in a distortion of the caregivers' view of the patients' sleep.

An important limitation of this study is its reliance on caregiver reports of sleep disturbance in the patients. Since studies have shown that caregiver depression, burden, and time spent caregiving may be associated with increased reporting of neuropsychiatric symptoms in dementia patients,³² we created regression models that took caregiver depression into account in evaluating the relation of ethnic status to sleep problems. As our routine clinical assessment did not include more detailed data on perceived or actual caregiving burden it was not possible to adjust analyses for this factor. Since at least one study has shown that the relation of caregiver reports of sleep problems to objective sleep measures is low,³¹ our results should be interpreted cautiously. Another important limitation is the lack of data on medication use in our data set. Other potential limitations in this study include the lack of information on patients' use of alcohol, caffeine, and tobacco, each of which might contribute to sleep disruptions. We cannot account for a possible confounding of our data by effects of assessor gender and ethnicity and the same characteristics among patients and caregivers. Finally, although our routine practice was to ask the person most familiar with the patient to accompany him or her to the evaluation and to serve as an informant, some caregivers did not actually live with the patients. Their reports of sleep problems may thus have been less reliable than those of caregivers who lived in the same home as did the patients.

These data thus show that sleep problems in patients with Alzheimer disease may be an important issue for clinicians to become aware of and to routinely assess. Although most significant differences implicated African Americans, it may be noted that in several cases average ratings for Hispanic patients were higher than those for non-Hispanic whites. As sleep problems may be a source of the disparities seen between minority and majority patients in health outcomes, they may be an important topic for additional research. These data might support the importance of developing culturally appropriate interventions for sleep disturbances in cognitively impaired African American elders, as well as provide support for the importance of educating the African American community on the importance of sleep as well as cognitive impairment. Educational efforts might also target clinicians who provide services to the African American community.

This study thus highlights the importance of assessing sleep disturbance in patients with dementia as well as the possibility that sleep disturbances may occur with greater severity in African Americans patients. As patient sleep disturbances may be especially stressful for caregivers, addressing these problems through clinical interventions may be a useful way to help caregivers maintain patients in the community. Since caregiver stress has been related to caregiver depression and mortality, supporting them is critically important. In addition, patient behavior problems are related to the likelihood that they will be institutionalized and thus are an important factor in health care costs. Sleep disturbance in patients with dementia is thus clinically significant and merits further investigation.

DISCLOSURE STATEMENT

This was not an industry supported study. The authors have indicated no financial conflicts of interest.