Relationship between behavioral disturbances and characteristics of patients in special units for dementia


Dr Keiko Matsuoka-Hattori, Yoshioka Rehabilitation Clinic, 4-2-8 Shimo-Renjaku, Mitaka-City, Tokyo 181-0013, Japan. Email:


Abstract  The aim of the present study was to examine the relationships between behavioral disturbances and patient characteristics of inpatients with dementia nationwide. The five patient characteristics used were age, gender, years of education, cognitive status, and walking ability. The subjects consisted of 730 inpatients selected by systematic sampling from 180 units that have specialized psychiatric beds for acute/long-term care of dementia. Clinical staff members assessed the 730 patients with the Mini-Mental State Examination (MMSE), and filled out a questionnaire for long-term care insurance, including 19 items relating to behavioral disturbances and walking scale. Five meaningful factors were identified out of the 19 behavioral disturbances by factor analysis. Linear regression analysis revealed that the factor ‘psychotic/neurotic’ was not related to any patient characteristics; ‘aggression/negativistic’ was related to male gender and a lower MMSE score; and ‘dirty/destructive’ and ‘disorientation/fire management’ were related to a lower MMSE score and higher walking score. The factor ‘sexual behavior’ included only one behavior at a very low frequency. These findings suggest that different behavioral disturbance factors have different correlations with patient characteristics, while cognitive dysfunction has a relatively important role in behavioral disturbances of inpatients with dementia.


Behavioral disturbances in patients with dementia are common, and are especially important for understanding the care burden of the family and clinical staff.1,2 Many previous studies have examined the etiological nature of such behaviors, especially their relationship to patient characteristics. For example, cognitive dysfunction (or dementia stage) was found to be associated with psychotic symptoms (i.e. delusions and/or hallucinations),3–7 agitation/aggression,8–13 and other behaviors such as wandering.9 In terms of gender, some studies reported a relationship between male gender and aggressive behavior.11,13,14 The length of education was also found to have an association with some behaviors.6,7,15 Although one study suggested the relationship between higher education and ‘frontal behavior’,6 lower educational level is generally related to some types of behavioral disturbances.7,15

However, most of these previous studies examined community dwellers, who had relatively mild levels of dementia  and  relatively  high  degrees  of  activity  of daily living (ADL). Behavioral disturbances are just as important for care in the clinical setting as for functionality in the community because behavioral disturbances often result in institutionalization, after which clinical care needs to cope with it. Little is known about the prevalence of behavioral disturbances and their association with patient characteristics in inpatients with dementia.

In the present study we aimed to clarify the prevalence of behavioral disturbances and their association with patient characteristics in a nationwide sample of patients hospitalized in specialized psychiatric beds for acute/long-term care of dementia.


Study population

The present study was conducted between December 1999 and March 2000, just before the launch of the new long-term care insurance system in Japan. This is a social insurance system for long-term care to meet the needs of elderly people, whose swelling numbers make Japan one of the most rapidly aging societies among the developed nations. In this new scheme, introduced in April 2000, care levels are classified into six categories, from ‘assistance required’ to care level 5 according to the time needed for care. Clients in care level 5 require the most time-consuming care. The primary care level is calculated automatically from the data of an 85-item questionnaire, while the final care level takes into account the patient's ADL level and comments by the patient's primary doctor.

In Japan, 248 units have specialized psychiatric beds for acute/long-term care for dementia at the time of investigation. We asked all of the units to participate in the present study; 180 (71.1%) units agreed. By systematic sampling, we selected five patients from each unit. Clinical staff members in charge of patients, such as doctors and nurses, performed assessments. Diagnoses of the subjects were made by psychiatrists according to the Diagnostic and Statistical Manual of Mental Disorders (4th edn; DSM-IV).16

The subjects in the present study consisted of 730 patients (81.1%: 730/900) with dementia whose scores on the 19-item behavior disturbance questionnaire (introduced later) and the Mini-Mental State Examination (MMSE) were available. Table 1 shows characteristics including primary assessment of care levels of the 730 patients as evaluated by the computer program. The mean age of the patients was 78.7 ± 8.8 years, with a range of 48–99 years. The mean educational level was 8.4 ± 3.0 years. Sixty-eight percent of the patients were female and 32% were male. Of the 730 patients, 81 (11.1%) were diagnosed as having early onset dementia of the Alzheimer type; 271 (37.1%) as having the late type; and 330 (45.2%) as having vascular dementia. The mean MMSE score was 9.2 ± 6.9).

Table 1. Characteristics of the 730 patients
  • Primary care level. Care level 1, elderly who requires assistance for 30–49 min a day; care level 2, for 50–69 min; care level 3, for 70–89 min; care level 4, for 90–109 min; care level 5, for ≥ 110 min.

  • MMSE, Mini-Mental State Examination.

Mean age (years)78.7 ± 8.8
Years of education  8.4 ± 3.0
Female subjects n (%)499 (68.4)
Dementia n (%)
 Alzheimer's type
   Early onset  81 (11.1)
   Late onset271 (37.1)
 Vascular dementia330 (45.2)
 Others  48 (6.6)
MMSE score  9.2 ± 6.9
Care leveln (%)
 Assistance required  62 (8.5)
 1157 (21.5)
 2147 (20.1)
 3187 (25.6)
 4125 (17.1)
 5  52 (7.1)

The questionnaire

The questionnaire included demographic characteristics, questions for eligibility assessment for long-term care insurance, and MMSE. The questions included 19 concrete examples of behavioral disturbances as follows: 1, feels persecuted, such as believing things are stolen; 2, makes up stories and tells them to other people; 3, sees and hears things that other people do not; 4, becomes emotionally unstable and cries and laughs a lot; 5, insomnia at night or reversal of day and night; 6, uses offensive language or becomes violent; 7, repeats the same story or makes an unpleasant noise; 8, shouts; 9, resists advice or care; 10, wanders; 11, becomes nervous such as saying ‘I am going home’; 12, once the patient goes out, he/she cannot find his/her way back to hospital, institution, or home; 13, the patient should be watched because he/she insists on going out alone; 14, collects various things or takes them without permission; 15, the patient cannot put out or deal with fire; 16, destroys or tears things or clothes; 17, dirty behavior (e.g. playing with feces); 18, puts inedible things into the mouth; and 19, sexual behavior that annoys other people.

These items are part of the primary care level assessment for long-term care insurance. Each item is rated on a 3-point scale of 1–3, with 1 indicating ‘never’ for that behavioral disturbance, 2 indicating ‘sometimes yes’, and 3 indicating ‘always yes’, by raters reviewing the latest 6 month behavioral records.

The questionnaire with these 19 behavioral disturbances is officially used in long-term care insurance in Japan, and is the most frequently used questionnaire for evaluating the amount of time required in care of the patients. However, the factor structure of these 19 behavioral disturbances has not been analyzed or reported. Therefore, we performed a factor analysis and varimax rotation on scores of the 19 behavioral disturbances to extract meaningful factors. We then calculated the total item scores of corresponding behavioral disturbances as the factor score. Using each of these factor scores as a dependent variable, we performed multiple regression analyses as described in the following section.

We used the walking scale assessed by the following criteria: 1, the patient can walk alone (if slowly) without any assistance; 2, the patient needs partial assistance for walking; and 3, the patient needs overall assistance for walking. This walking scale is included in the long-term care insurance questionnaire as well as the 19 behavioral disturbances, although it is not used for evaluating the primary care level.

In the analysis we used five patient variables as independent variables for multivariate regression analysis as follows: current age, gender (1, male; 2, female), years of education, MMSE score, and walking score. The reasons we employed these five variables are as follows. Several studies have reported that older patient age was often related to psychotic behavior such as delusions,3,7,17 whereas other studies did not find such a relationship.15,18 The age effect on behavioral disturbances in dementia is still unclear. As noted in the introduction, the variables of gender, education, and cognitive stage have been examined but their relationship with behavioral disturbances remains controversial. And, in spite of its importance, little is known about the relationship between walking ability and behavioral disturbances. We hypothesized that walking ability might tend to increase behavioral disturbance in inpatients whose dementia is relatively severe.

Because no correlation between any two of these five variables exceeded 0.35 (Pearson's correlations between parametric variables; Spearman's correlations between non-parametric variables), all five variables were entered in multiple linear regression analyses. All statistical analyses were carried out with spss for Windows, version 10.0J (SPSS Japan, Tokyo, Japan).


The internal consistency reliability (Cronbach's α) of 19 behavioral disturbances was acceptably high (α = 0.84) in our sample.

Table 2 shows the prevalence of the 19 behavioral disturbances and the results of factor analysis. Prevalence varied from 5.3% (‘sexual behavior that annoys other people’) to 65.1% (‘once the client goes out, he/she cannot find his/her way back to hospital, institution, or home’). The factor analysis yielded five factors with eigenvalues >1 (explained variance: 57.9%). Each of these factors represented clinically interpretable domains. Factor 1 represented ‘psychotic/neurotic’ (explained variance: 14.4%), factor 2 represented ‘aggression/negativistic’ (explained variance: 13.1%), factor 3 represented ‘dirty/destructive’ (explained variance: 13.0%), factor 4 represented ‘disorientation/fire management’ (explained variance: 11.5%), and factor 5 was ‘sexual behavior’ (explained variance: 5.9%). Because factor 5 ‘sexual behavior’ included only one item, we excluded this factor from further analysis.

Table 2.  Factor loadings of the behavioral disturbances in the 730 patients
Behavioral disturbance used in long-term care insurance% of answer
2 (sometimes yes)/
3 (always yes)
Factor 1: psychotic/neurotic (α = 0.74) 2.7414.4 
 Feels persecuted such as believing things are stolen39.3  0.76
 Makes up stories and tells them to other people37.7  0.76
 Sees and hears things that other people do not36.3  0.64
 Repeats the same story or makes an unpleasant noise35.3  0.56
 Becomes nervous such as saying ‘I am going home’40.1  0.56
 Insomnia at night or reversal of day and night45.6  0.35
Factor 2: aggression/negativistic (α = 0.78) 2.4813.1 
 Uses offensive language or becomes violent38.8  0.82
 Shouts32.6  0.81
 Resists advice or care53.7  0.76
 Becomes emotionally unstable and cries and laughs a lot43.9  0.50
Factor 3: dirty/destructive (α = 0.71) 2.4813.0 
 Dirty behavior (e.g. playing with feces)39.5  0.72
 Puts inedible things into mouth14.5  0.71
 Destroys or tears things or clothes18.1  0.65
 Wanders53.7  0.60
 Collects various things or takes them without permission25.9  0.57
Factor 4: disorientation/fire management (α = 0.76) 2.1911.5 
 Once the client goes out he/she cannot find his/her way   back to hospital, institution, or home65.1  0.83
 The client cannot put out or deal with fire46.4  0.82
 The client should be watched because he/she insists on   going out alone48.1  0.73
Factor 5: sexual behavior 1.12  5.9 
 Sexual behavior that annoys other people  5.3  0.91

Table 3 shows the results of multiple regression analyses. The five variables explained only 1.4% of the factor variance for ‘psychotic/neurotic’; none of the adjusted beta variables reached a significant level. For the factor ‘aggressive/negativistic’, the five variables explained 8.3% of the variance, with gender (adjusted beta = −0.153, P < 0.05) and MMSE score (adjusted beta = −0.183, P < 0.01) being significant. For the factor ‘dirty/destructive’, the five variables explained 20.9% of the factor variance, with MMSE score (adjusted beta = −0.489, P < 0.01) and walking score (adjusted beta = −0.262, P < 0.01) reaching a significant level. For the factor ‘disorientation/fire management’, the five variables explained 9.3% of the variance, with MMSE score (adjusted beta = −0.160, P < 0.05) and walking score (adjusted beta = −0.328, P < 0.01) reaching a significant level.

Table 3. Linear regression analyses with each factor score as a dependent variable
Independent variablesFactor: adjusted beta
fire management
  • MMSE, Mini-Mental State Examination.

  • *

    P  < 0.05;

  • * *

    P  < 0.01.

Age (years)  0.083 −0.101 −0.007  0.044
Gender (male = 1, female = 2)  0.035 −0.153*  0.028  0.049
Years of education −0.002 −0.081  0.123  0.067
MMSE score −0.030 −0.183** −0.489** −0.160*
Walking score −0.102  0.102 −0.262** −0.328**
Adjusted R2  0.014  0.083  0.209  0.093


Not surprisingly, the prevalence of each behavioral disturbance in patients occupying specialized psychiatric beds for acute/long-term care of dementia was almost always higher than that of community dwellers.19 For example, the prevalence of ‘wandering’ is reportedly 9% among patients with dementia living in Tokyo and 53.7% among inpatients in our study.19 This result indicates that behavioral disturbances are strongly related to hospitalization (i.e. important causes of hospitalization). Therefore, reduction of behavioral disturbances, not of dementia itself, is an important role of special beds for dementia.

We extracted five factors (‘psychotic/neurotic’, ‘aggression/negativistic’, ‘dirty/destructive’, ‘disorientation/fire management’, ‘sexual behavior’) from 19 behavioral disturbances used in the assessment of eligibility for long-term care insurance; these factors were differently correlated with patient characteristics.

Contrary to our assumption, the factor ‘psychotic/neurotic’ was not related to any patient characteristics. This result differed from some previous studies in which patients who presented with psychotic symptoms had lower cognitive function.3–7 One reason for this disagreement might be the relationship between psychotic/neurotic behaviors and cognitive status. As some studies have suggested,12,15,20–22 such relationships might be curvilinear rather than linear; namely, peaking in the moderate stage and then declining in the severe stage. In other words, our linear regression analyses might have been inappropriate for examining a curvilinear relationship.

The factor ‘aggression/negativistic’ was the only variable that related to gender; male subjects tended to show a higher aggression/negativistic score. This result is congruent with those reported by Lyketsos et al. and Jagger and Lindesay.11,14 Additionally, lower MMSE scores were also associated with higher aggression/negativistic scores, in accordance with many previous studies.8–13 These findings suggest that premorbid personality traits may be intensified (or exaggerated) by cognitive dysfunction because male gender is more often related to aggressive personality traits than female gender.

The factor ‘dirty/destructive’ and ‘disorientation/fire management’ had significant relationships with MMSE score and walking score. Interestingly, lower MMSE and higher walking ability contributed to these factor scores. Several studies have pointed out that behavioral disturbances generally peak in the moderate to severe stage, but not in the most severe stage.12,15,20,21 As our results suggested, these findings might be explained in terms of the gap between cognitive dysfunction and walking level.

In relation to the MMSE score, we assume the factors aggression/negativistic, dirty/destructive, and disorientation/fire management reflected some kind of brain damage, such as frontal dysfunction. However, although many studies have examined neurological and biological correlates of psychotic symptoms in patients with dementia, few have addressed neurobiological correlates with the behavioral symptoms dirty/destructive and disorientation/fire management. Further study is needed to clarify how cognitive dysfunction causes such severe problem behaviors.

Of the five factors, ‘sexual behavior’ consisted of only one questionnaire item (i.e. sexual behavior that annoys other people). The frequency of this behavior was relatively low in our inpatient sample (only 5% were rated as sometimes yes or always yes), suggesting that this behavior had a different cause from the list of behavioral disturbances in the inpatient sample. The answer to the question about sexual behavior seemed to depend on the caregiver's subjective feeling toward the sexual-like behavior. In clinical settings, unlike community settings, sexual-like behavior might not be regarded as problematic by professional, trained caregivers. In that case, the low prevalence of sexual behavior in patients with dementia in clinical settings does not necessarily mean a low prevalence of such behavior in community-dwelling patients. Further study is needed to clarify the prevalence and correlates of sexual behavior in community-dwelling patients with dementia.

Several concerns must be raised regarding these results. First, variables used in the present study were too small to explain behavioral disturbances. It must be noted that R2 values in the present study were not high, especially the factor ‘psychotic/neurotic’. The present study did not include items on delirium (often a cause of some kinds of behavioral disturbances), on history of alcoholism, or on experience of stroke. Second, our data were cross-sectional, and could not add insights into valid longitudinal changes in each patient.

In conclusion, the present findings suggest that different behavioral disturbance factors have different correlations with the characteristics of inpatients with dementia. Cognitive dysfunction measured by MMSE contributed to three factors of behavioral disturbances: ‘aggression/negativistic’, ‘dirty/destructive’, and ‘disorientation/fire management’, suggesting that cognitive dysfunction is a relatively important correlate for behavioral disturbances in such patients.


We would like to thank Dr M. Uno, the director of Yoshioka Rehabilitation Clinic, and F. Nakahara, the head of care worker in Taro Geriatric Health Services Facility, for their helpful advice.