The Minimum Data Set Agitated and Reactive Behavior Scale (MDS ARBS): Measuring Behaviors in Nursing Home Residents with Dementia

Abstract

Objectives:

Interventions aimed at managing agitated and aggressive behaviors in dementia without the use of antipsychotics are currently being tested in nursing homes (NHs). Researchers and clinicians require a measure that can capture the severity of residents’ behaviors. We test the internal consistency and construct validity of the Agitated and Reactive Behavior Scale (ARBS), a measure created using data from mandatory NH assessments.

Design:

Cross-sectional

Setting and Participants:

2016 national sample of 15,326 CMS-certified NHs. 489,854 new admissions and 765,367 long-stay residents (at least 90 days in NH). All participants have a dementia diagnosis.

Methods:

Minimum Data Set (MDS), version 3.0. The ARBS is a composite measure of: 1) physical behavioral symptoms directed at other people; 2) verbal behavioral symptoms directed at other people;3) other behavioral symptoms not directed at other people; and 4) rejection of care. Variables used to establish construct validity included: degree of cognitive impairment, use of medications for managing agitation and aggression, and co-occurring conditions associated with agitated and aggressive behaviors (e.g., schizophrenia, depression, or delirium).

Results:

This paper has three important findings: 1) the ARBS score has borderline-adequate internal consistency (α=.64-.71) in the national population NH residents with dementia; 2) only 18% of new admissions and 21% of long-stay residents with dementia evidence any agitated or aggressive behaviors in the last week, as rated in the MDS assessment; and, 3) the ARBS demonstrates good construct validity; it increases with cognitive impairment, treatment with relevant medications, and co-occurring psychiatric conditions and symptoms.

Discussion:

Nationally available MDS data may significantly underestimate the prevalence of agitated and aggressive behaviors among NH residents with dementia.

Conclusions and Implications:

Researchers conducting pragmatic trials of non-pharmaceutical interventions to manage behaviors in NH residents with dementia should consider the likely under-detection of these behaviors in the available MDS data.

Brief summary:

Mandatory assessments of nursing home residents likely under-detect the frequency of agitated and aggressive behaviors which has implications for pragmatic evaluations of non-pharmacologic interventions.

Introduction

Approximately half of all people living in nursing homes (NHs) have dementia,^{1, 2} and between 48 to 82 percent of NH residents with dementia exhibit agitated or aggressive behaviors.³ Many of these behaviors are directed toward NH staff who assist with bathing, dressing, and feeding residents.⁴ Growing attention to the risks associated with off-label use of antipsychotic medications, including increased risk of falls⁵ and death,^{6, 7} has decreased their popularity as a first-line treatment for managing agitated and aggressive behaviors in dementia.^8–10 Yet, agitated and aggressive behaviors persist. Finding non-pharmaceutical techniques to manage behaviors is critical to the safety and well-being of NH residents and staff.

As non-pharmacological interventions to manage agitated and aggressive behaviors are tested in NHs, researchers and clinicians need a measure that can systematically capture the effects of these interventions on these behaviors. The Aggressive Behavior Scale (ABS), a summary scale developed by Perlman and Hirdes in 2008, may be one such measure.¹¹ The ABS is easily constructed from data routinely collected for every resident in Centers for Medicare & Medicaid Services (CMS)-certified NHs. This allows for pragmatic evaluation of non-pharmaceutical alternatives, across multiple NHs, without prohibitive data collection costs.

In the ten years since the ABS was developed, thinking about agitated and aggressive behaviors in dementia, and the causes of these behaviors, has evolved. Although not reflected in the name, the ABS is comprised of behaviors not directed toward others (often classified as agitated behaviors) and behaviors directed toward others (often classified as aggressive behaviors). Agitation and aggression are distinct states with different etiologies; labeling a resident as aggressive can be stigmatizing, as most aggressive behaviors in dementia are the result of a reaction to an external stimuli (e.g., staff touching resident for care activities).¹² For these reasons, we have renamed the updated scale the Agitated and Reactive Behaviors Scale (ARBS).

The ABS was developed and validated on a Canadian sample using the Resident Assessment Instrument –Minimum Data Set (RAI-MDS), version 2.0.¹³ Using similar methods to the original report, we report the internal consistency and construct validity of the ARBS using RAI-MDS data, version 3.0. This is the first time the ARBS has been validated for all US NH residents with dementia.

Methods

Data & Sample

The clinical information, cognitive status, physical function, and personal preferences of residents of CMS-certified NHs are assessed upon admission and at least once per quarter using a standardized assessment tool, the RAI. Although the quarterly assessments are abbreviated, they contain all items necessary for constructing the ARBS. Data from these assessments are referred to as the MDS. RAI-MDS, version 3.0, was released on October 1, 2010.

The study samples include: 1. Newly admitted residents with dementia (admitted to CMS-certified NH during 2016), and 2. Long-stay residents with dementia (at least 90 days in NH, as evidence by a quarterly or annual assessment during 2016). Most of the newly admitted residents are short-stay, or rehabilitation patients, entering after a hospital stay. However, there are some new admissions which enter from the community in the new admission sample. To be included in the new admission sample, residents must not have been admitted to a nursing home in the past year and must have a 14-day assessment. Residents with Alzheimer’s Disease or other dementias (includes Lewy Body, vascular, or multi-infarct) were identified using active diagnosis checkboxes or related ICD-10 codes.

Variables

Agitated and Reactive Behavior Scale (ARBS)

The ARBS is a composite measure that describes the frequency of agitated and reactive behaviors.¹¹ The four items that compose the ARBS include: 1. physical behavioral symptoms directed at other people (e.g., hitting, kicking, pushing) (MDS Item E0200A); 2. verbal behavioral symptoms directed at other people (e.g., threatening, screaming, or cursing) (MDS Item E0200B); 3.other behavioral symptoms not directed at other people (e.g., pacing, rummaging, disrobing) (MDS Item E0200C); and 4. resident rejection of necessary care (MDS Item E0800). Response categories for all four items are: behavior was not exhibited in the last week (0), behavior occurred 1–3 days (1), behavior occurred 4–6 days (2), or behavior occurred daily (3). These four items are summed to create the ARBS, which takes on values from 0 (behavior not exhibited on all four items) to 12 (behavior occurred daily on all four items).

Cognitive Impairment

We used the Cognitive Function Scale (CFS), to classify residents’ cognitive functioning as: cognitively intact, mildly impaired, moderately impaired, or severely impaired.¹³ The CFS is created using the Brief Interview for Mental Status (BIMS) and the Cognitive Performance Scale (CPS). The BIMS is a resident screening tool. The CPS is used to account for missing data on the BIMS.

Other Construct Validity Variables

We considered the relationship between ARBS severity and medications used to manage agitation and aggression in the last seven days, including antipsychotics, antianxietals, and antidepressants. We also considered the relationship between ARBS severity and scheduled and PRN pain medications used in the past five days.

We considered the relationship between the ARBS and other co-occurring conditions that may result in agitated and aggressive behaviors, including delirium, presence of hallucinations or delusions, frequent pain, schizophrenia, psychotic disorder (not schizophrenia), bipolar disorder, and depression. Signs and symptoms of delirium are identified by the Confusion Assessment Method.¹⁴ Frequent pain is defined as resident-report of frequent or constant pain or staff report of likely pain in 3 of last 5 days. Schizophrenia, psychotic disorder (not schizophrenia), bipolar disorder, and depression, were identified using diagnostic checkboxes.

Analysis

Internal consistency of the ARBS was tested using Cronbach alpha coefficients. Construct validity was assessed as differences in ARBS severity across related constructs. All analyses were conducted using STATA, SE Version 14. Per RECORD guidelines, replication information can be found in our repository (https://repository.library.brown.edu/studio/item).

Results

Samples

The analytic sample included 489,854 new admissions and 765,367 long-stay NH residents with dementia in 2016.(Table 1) Despite all having a dementia diagnosis, 20% of new admissions were cognitively intact based on the CFS, 26% had mild impairment, 44% had moderate impairment, and 9% were severely impaired. Among the long-stay population with a dementia diagnosis, 14% of the long-stay population was cognitively intact based on the CFS, 22% had mild impairment, 47% had moderate impairment, and 17% were severely impaired. Antidepressant use was common among new admissions and long-stay residents with dementia (46% and 56%, respectively), as was antipsychotic use (24% and 25%, respectively), and antianxietal use (19% and 22%, respectively). PRN pain medications were common among new admissions (44%) and scheduled pain medications were common among long-stay residents (40%). Over half (53%) of long-stay residents and 38% of new admissions with dementia were depressed. Frequent pain was relatively uncommon among new admissions and long-stay residents (12% and 6%, respectively).

Table 1.

New admissions and long-stay nursing home residents with dementia diagnosis in 2016

	New Admissions (N= 489,854)		Long-Stay Residents (N=765,367)
	N	%	N	%
Behaviors in the past week
Any physical behaviors directed toward others	31,463	(6.4)	48,533	(6.3)
Any verbal behaviors directed toward others	35,901	(7.3)	71,846	(9.4)
Any other behaviors not directed toward others	32,220	(6.6)	62,642	(8.2)
Any rejection of care in past week	53,619	(11.0)	81,688	(10.7)
Demographics
Age, Mean(SD)	83.3 (9.1)		83.3 (10.3)
Male	182,506	(37.3)	230,592	(30.1)
Black	51,920	(10.6)	100,836	(13.2)
Hispanic	24,212	(4.9)	43,197	(5.6)
Dementia Diagnosis
Alzheimer’s Dementia	109,094	(22.3)	213,252	(27.9)
Non-Alzheimer’s Dementia	428,622	(87.5)	669,768	(87.5)
Cognitive Function
Intact	96,700	(19.7)	109,315	(14.3)
Mild	129,399	(26.4)	165,032	(21.6)
Moderate	215,079	(43.9)	355,770	(46.5)
Severe	44,102	(9.0)	129,912	(17.0)
Missing	4,574	(0.9)	5,338	(.7)
Drugs (Any use)
Antipsychotics (last 7 days)	116,910	(23.9)	190,536	(24.9)
Antianxiety (last 7 days)	92,418	(18.9)	164,363	(21.5)
Antidepressants (last 7 days)	225,973	(46.1)	430,630	(56.3)
Scheduled Pain Medication (last 5 days)	132,119	(27.0)	307,252	(40.1)
PRN Pain Medication (last 5 days)	214,122	(43.7)	166,518	(21.8)
Other Conditions & Symptoms
Schizophrenia	12,701	(2.6)	48,406	(6.3)
Psychotic Disorder (not Schizophrenia)	27,507	(5.6)	101,608	(13.3)
Bipolar Disorder	12,378	(2.5)	34,562	(4.5)
Depression	184,743	(37.7)	406,911	(53.2)
Hallucinations or Delusions	24,888	(5.1)	61,251	(8.0)
Delirium	34,624	(7.1)	97,575	(12.8)
Frequent Pain	60,435	(12.3)	46,477	(6.1)

Relationship between agitated and aggressive behaviors and degree of cognitive impairment

Internal score consistency of the ARBS was .71 for new admissions and .65 for long-stay residents with dementia. When residents with co-occurring schizophrenia, bipolar disorder, or intellectual disability were excluded, the internal score consistency of the ARBS was .71 for new admissions and .66 for long-stay residents. The vast majority of new admissions (82%) and long-stay residents (79%) with dementia had no agitated or aggressive behaviors as measured by the ARBS (ARBS=0). The percent of residents with any agitated or aggressive behaviors in the past week increased by degree of cognitive impairment.(Figure 1) Only 8% of cognitively intact new admissions had any agitated or aggressive behaviors, compared to 13% of new admissions with mild impairment, 23% with moderate impairment, and 31% with severe impairment. Similarly for long-stay residents, 13% of cognitively intact residents had any behaviors, compared to 16% of residents with mild impairment, 24% with moderate impairment, and 26% with severe impairment.

Figure 1.

Percent of new admissions and long-stay nursing home residents with a dementia diagnosis who had any agitated or aggressive behaviors in the past week by degree of cognitive impairment. Data from all nursing homes certified by the Centers for Medicare and Medicaid Services during 2016.

Using the categories of the original ABS established by Ahn et al. (2013),¹⁵ we find the ARBS severity was also related to the degree of cognitive impairment.(Table 2). Among new admissions with mild to moderate behaviors (ARBS=1,2), 56% had moderate cognitive impairment and 13% had severe impairment. Among new admissions with very severe behaviors (ARBS>=6), 58% had moderate cognitive impairment and 29% had severe cognitive impairment. A similar relationship between ARBS severity and degree of cognitive impairment was observed for long-stay residents.

Table 2.

Relationship between severity of ABS and related constructs for new admissions and long-stay nursing home residents with a dementia diagnosis

	New Admissions (N= 489,854)				Long-Stay Residents (N=765,367)
	No Behaviors (ABS=0)	Mild to Moderate Behaviors (ABS 1–2)	Severe Behaviors (ABS 3–5)	Very Severe Behaviors (ABS ≥6)	No Behaviors (ABS=0)	Mild to Moderate Behaviors (ABS 1–2)	Severe Behaviors (ABS 3–5)	Very Severe Behaviors (ABS ≥6)
(Residents, N)	(397,199)	(60,442)	(22,508)	(5,131)	(597,933)	(109,688)	(41,865)	(10,543)
	Column %				Column %
Degree of Cognitive Impairment
Intact	22.4	10.2	5.5	3.1	16.0	9.7	6.6	3.6
Mild	28.3	21.8	15.2	9.5	23.1	18.4	13.5	9.5
Moderate	41.6	55.5	59.9	58.0	44.9	53.2	55.0	53.9
Severe	7.7	12.6	19.4	29.3	16.0	18.7	24.9	33.0
Medication Use
Antipsychotics*	20.0	36.3	49.5	60.9	21.1	35.8	44.2	50.0
Antianxiety*	16.1	27.5	37.8	45.9	18.5	30.0	36.5	40.1
Antidepressants*	45.3	49.4	50.9	53.0	55.0	61.3	60.8	59.8
Scheduled Analgesic†	26.9	27.3	27.4	28.0	39.7	42.0	41.9	42.4
PRN Analgesic†	44.1	42.7	41.8	39.6	21.4	23.5	23.0	22.2
Other Conditions & Symptoms
Schizophrenia	2.3	3.6	4.5	5.2	5.6	8.1	10.3	11.3
Other Psychotic Disorder	4.6	8.7	12.3	15.4	11.6	18.2	21.5	24.3
Bipolar Disorder	2.3	3.2	3.6	4.4	4.1	5.7	6.6	7.1
Depression	37.4	39.3	39.1	38.9	52.4	56.4	55.7	54.7
Hallucinations / Delusions	2.7	12.0	21.9	32.6	4.9	15.7	25.4	34.4
Delirium	7.4	17.5	27.0	33.2	10.3	19.5	26.0	31.0
Frequent Pain	12.1	13.0	13.8	14.0	5.7	7.1	7.7	7.9

^*Any use in last 7 days

^†Any use in last 5 days

Relationship between ARBS and other potentially related constructs

Drugs used to treat agitated and aggressive behaviors increased with ARBS severity (Table 2). The strongest relationships were found for antipsychotics and antianxietals. For new admissions, 20% of those with no behaviors (ARBS=0) received an antipsychotic in the last week, compared to 36% of new admissions with mild to moderate behaviors (ARBS=1,2), 50% of new admissions with severe behaviors (ARBS=3,4,5), and 61% of new admissions with very severe behaviors (ARBS>=6). A similar relationship between antipsychotic use and ARBS severity was observed among long-stay NH residents, as 21% of residents with no behaviors received an antipsychotic in the past week, compared to 50% of residents with very severe behaviors. Similarly, antianxietal use increased with ARBS severity: 16% of new admissions with no behaviors received antianxietals, compared to 28% of new admissions with mild to moderate behaviors, 38% of new admissions with severe behaviors, and 46% of residents with very severe behaviors. Among long-stay residents with dementia, 19% of residents with no behaviors received an antianxietal in the last week, compared to 40% of residents with very severe behaviors. A lesser relationship was observed between ARBS severity and antidepressant use. Among new admissions with no behaviors, 45% received an antidepressant, compared to 53% of new admissions with severe behaviors (55% of long-stay with no behaviors, 60% of long stay with severe behaviors). Scheduled and PRN analgesic use did not significantly increase with ARBS severity.

Co-occurring serious mental illness also increased with ARBS severity: 2% of new admissions with no behaviors had schizophrenia, compared to 5% of new admissions with very severe behaviors; 6% of long-stay residents with no behaviors had schizophrenia, compared to 11% of residents with very severe behaviors. Similar patterns were found for bipolar disorder (2% of new admissions with no behaviors, 4% with severe behaviors; 4% of long-stay with no behaviors, 7% with severe behaviors), and other psychotic disorders (5% of new admissions with no behaviors, 15% with severe behaviors; 12% of long-stay with no behaviors, 24% of long-stay with severe behaviors). Presence of delirium, delusions and hallucinations, and frequent pain were also associated with ARBS severity.

Discussion

This paper has three important findings. First, the ARBS score has borderline-adequate internal consistency (α=.71 for new admissions, α=.64 for long-stay residents) in the national population NH residents with dementia. Second, only 18% of new admissions and 21% of long-stay residents with dementia have any agitated or aggressive behaviors in the last week, as captured by the MDS assessment. The ARBS demonstrates construct validity by increasing with degree of cognitive impairment, use of medications used to control agitated and aggressive behaviors, and co-occurring psychiatric conditions and symptoms.

It is likely that the ARBS under-detects agitated and aggressive behaviors. The prevalence of agitated and aggressive behaviors identified by the ARBS (approximately 21%) is less than the existing literature would suggest (between 48 to 82 percent).³ The items that comprise the ARBS are derived from the Cohen-Mansfield Agitation Inventory (CMAI), a validated tool used to interview staff about resident behaviors.^16–18 During the development of MDS, version 3.0, Saliba et al. (2008) validated the behavioral measures in the MDS against the CMAI (gold standard measure) in a small sample of nursing home residents and found that, while related MDS and CMAI domains demonstrated agreement, the number of behaviors detected by the MDS were approximately half the number of behaviors detected by the CMAI.¹⁹ Bharucha et al. (2008) found similar levels of underdetection.²⁰ Under-detection of behaviors and conditions in the MDS is common, due to staff normalization of behaviors and lack of knowledge of residents by staff completing assessments.^21–24 Unfortunately, CMAI data is not available for all nursing home residents.

Despite under-detection, the ARBS demonstrates construct validity. We find high levels of antipsychotic use among residents with dementia and severe scores on the ARBS (over half of new admissions and long-stay residents with very severe behaviors received antipsychotics in the past week). We also observed relationships between psychiatric diagnoses and the ARBS, including bipolar disorder, schizophrenia, other psychosis, and delirium. While bipolar disorder and schizophrenia are relatively rare diagnoses, other psychosis (not schizophrenia) was relatively common (13% of long-stay NH residents with dementia). While out of the scope of this paper, it may be worth considering changes in prevalence of “other psychosis” over time. Increased use of exclusionary diagnoses, in response to pressures to reduce off-label use of antipsychotics in dementia, has been detected.^25,26

Higher ARBS scores were associated with more frequent pain. Other work has found agitated and aggressive behaviors in dementia may be the result of under-detected and undertreated pain.¹⁴ We also found a subtle relationship between antidepressant use and ARBS. Recent work has suggested antidepressants are being used as alternatives to antipsychotics to manage these behaviors.⁸ These are two areas ripe for further research.

Limitations

There are some limitations to using Cronbach’s alpha as a measure of internal consistency, namely that the resulting coefficient can be manipulated by increasing the number of items comprising the scale (only four items in the ARBS), and inter-relatedness of items does not imply unidimensionality.²⁷ The focus of this work was to create a simple scale that characterizes residents’ agitated and reactive behaviors and demonstrates construct validity. Depending on the intervention being tested, researchers may want to focus on the items of the scale that are only related to agitation or only related to reactive behaviors; factor or latent class analyses may be helpful in those cases. This is a point in time assessment of the relationship between the ARBS, cognitive impairment, and other related constructs. Future work is needed to understand the sensitivity of the ARBS to change over time.²⁸

Conclusions and Implications

The internal consistency for the ARBS is borderline adequate, and under-detection of behaviors is likely. Researchers conducting pragmatic trials of non-pharmaceutical interventions to reduce agitated and aggressive behaviors in NH residents with dementia should consider the likely under-detection of these behaviors in the available MDS data.