The Impact of Agitation in Dementia on Caregivers: A Real-World Survey

INTRODUCTION

In patients with dementia, cognitive, functional, and behavioral impairments lead to reduced patient autonomy. These symptoms often lead to dependence on non-professional or professional caregivers [1, 2]. According to analysis of data from the 2011 United States (US) National Health and Aging Trends Study, around 80% of patients with dementia required help of non-professional caregivers [3].

Agitation is a behavioral symptom of dementia. Behavioral or neuropsychiatric symptoms have been reported to be particularly burdensome to caregivers of dementia patients [4, 5], with disruptive behaviors such as agitation, aggression, and disinhibition presenting the greatest burden [6]. There is no widespread agreement on the elements involved in aggression, but the International Psychogeriatric Association (IPA) has published a provisional consensus on a definition of agitation in cognitive disorders, which includes a specification of ‘behaviors that cause excess disability’ [7]. Likely due at least in part to the uncertainty around what constitutes agitation, wide ranges in the prevalence of this symptom in dementia have been reported in the literature. A systematic review of studies of behavioral disturbances in dementia focused on agitation and included studies based on the IPA consensus definition; it included studies that did not explicitly use the term agitation but reported on behaviors included in the IPA definition [8]. This review reported the prevalence of agitation to range from 5% to 88% across all studies, with 21 (38%) studies reporting a prevalence of agitation of ≥50% [8].

The American Psychiatric Association Practice Guidelines advocate a personalized approach to the treatment of agitation in dementia, with antipsychotic drugs recommended when agitation symptoms are severe; however, there is no US Food and Drug Ad-ministration (FDA)-approved pharmacologic treatment for agitation in dementia [9].

The primary objective of this survey was to assess the association of agitation severity with the number of non-professional caregiver hours and non-professional caregiver burden, health status, and productivity for all patients, regardless of whether this caregiving was defined as proximate care (non-professional caregiver living with the patient) or remote care (non-professional caregiver not living with the patient). A secondary objective was to assess the association of agitation severity with these outcomes separately for patients receiving either proximate care or remote care.

MATERIALS AND METHODS

RESULTS

Impact of agitation for all patients

Caregiver hours

Regression analysis showed that the relationship between the number of hours per week of non-professional caregiving and increasing agitation severity score was positive, but non-significant, when all patients were included in the analysis (p = 0.314; Fig. 1). Regression results including IRR and 95% CI can be found in Supplementary Figure 1.

Fig. 1

Negative binomial regression analysis of caregiver hours by agitation score –all patients. An IRR significantly different from 1 indicates caregiver hours differ significantly for different agitation scores; an IRR > 1 implies greater hours with higher agitation scores, an IRR < 1 implies fewer hours with higher agitation scores. All regressions were controlled for patient demographics (age and sex) and clinical characteristics (time since diagnosis, current MMSE score). Coef, coefficient; CI, confidence interval; IRR, incidence rate ratio; MMSE, Mini-Mental State Examination.

Caregiver burden

Regression analyses showed ZBI-assessed Total Caregiver Burden, Personal Strain, Role Strain, and Guilt all increased with increasing agitation severity score when the analyses considered all patients, regardless of whether receiving remote or proximate caregiving (all p < 0.01, except Guilt p < 0.05; Fig. 2). The association with agitation severity was most marked for Total Caregiver Burden, and Personal Strain. Regression results including coefficients and 95% CI can be found in Supplementary Figure 2.

Fig. 2

Linear regression analysis of caregiver burden assessed with ZBI by agitation score –all patients. A regression coefficient significantly different from 0 indicates the predicted outcomes differs significantly for different agitation scores; a coef > 0 implies a worse outcome with higher agitation scores, a coef < 0 implies a better outcome with higher agitation scores. All regressions were controlled for patient demographics (age and sex) and clinical characteristics (time since diagnosis, current MMSE score). Coef, coefficient; CI, confidence interval; MMSE, Mini-Mental State Examination; ZBI, Zarit Burden Interview.

Caregiver health status

Caregivers’ EQ-5D-3L health utility index and EQ-5D-3L VAS were negatively (indicating poorer health status), but non-significantly, associated with increasing agitation severity (EQ-5D-3L health utility index regression coefficient [95% CI]: –0.003 [–0.009, 0.003]; p = 0.294; EQ-5D-3L VAS score regression coefficient [95% CI]: –0.036 [–1.09, 0.27]; p = 0.259). The regression plots can be found in Supplementary Figure 3.

Caregiver productivity

WPAI-assessed presenteeism, overall work im-pairment and total activity impairment all increased with increasing agitation severity score based on regression analyses (p < 0.01, p < 0.05 and p < 0.001, respectively; Fig. 3). Absenteeism was negatively, but non-significantly, associated with increasing agitation severity (Fig. 3). The regression plots can be found in Supplementary Figure 4.

Fig. 3

Linear regression analysis of WPAI-assessed productivity and activity by agitation score –all patients. A regression coefficient significantly different from 0 indicates the predicted outcomes differs significantly for different agitation scores; a coef > 0 implies a worse outcome with higher agitation scores, a coef < 0 implies a better outcome with higher agitation scores. All regressions were controlled for patient demographics (age and sex) and clinical characteristics (time since diagnosis, current MMSE score). Coef, coefficient; CI, confidence interval; MMSE, Mini-Mental State Examination; WPAI, Work Productivity and Activity Index.

Impact of agitation dependent on type of care

Caregiver hours

The number of non-professional caregiver hours/week was higher for patients receiving proximate care than remote care, at all agitation severity score levels. Regression analysis showed a significant increase in the number of hours per week of caregiving with increasing agitation severity score for patients receiving proximate care (p = 0.001; Fig. 4), but not for those receiving remote care (IRR [95% CI]: 1.03 [0.92, 1.16]; p = 0.609). The regression plots can be found in Supplementary Figure 5.

Fig. 4

Linear regression analysis of caregiver hours by agitation score and type of care. Proximate Care, caregiver living with patient; Remote Care, caregiver not living with patient. An IRR significantly different from 1 indicates caregiver hours differ significantly for different agitation scores; an IRR > 1 implies greater hours with higher agitation scores, an IRR < 1 implies fewer hours with higher agitation scores. All regressions were controlled for patient demographics (age and sex) and clinical characteristics (time since diagnosis, current MMSE score). Coef, coefficient; CI, confidence interval; IRR, incidence rate ratio; MMSE, Mini-Mental State Examination.

Caregiver burden

Based on the findings from regression analyses, ZBI Total Caregiver Burden, Personal Strain, and Role Strain all increased with increasing agitation severity score for caregivers providing proximate care (p < 0.01, p < 0.01, and p < 0.001, respectively; Fig. 5) but not for caregivers providing remote care (all p > 0.05; Fig. 5). Regression analyses showed an association of increasing ZBI Guilt with increasing agitation severity score for caregivers providing remote care (p < 0.001; Fig. 5), but not for caregivers providing proximate care (p > 0.05; Fig. 5). The regression plots can be found in Supplementary Figure 6.

Fig. 5

Linear regression analysis of caregiver burden assessed with ZBI by agitation score and type of care. Proximate Care, caregiver living with patient; Remote Care, caregiver not living with patient. A regression coefficient significantly different from 0 indicates the predicted outcomes differs significantly for different agitation scores; a coef > 0 implies a worse outcome with higher agitation scores, a coef < 0 implies a better outcome with higher agitation scores. All regressions were controlled for patient demographics (age and sex) and clinical characteristics (time since diagnosis, current MMSE score). Coef, coefficient; CI, confidence interval; MMSE, Mini Mental State Examination; ZBI, Zarit Burden Interview.

Caregiver health status

Regression analyses did not show a statistical association of caregivers’ EQ-5D-3L health utility index or VAS score with agitation severity score for caregivers providing either remote or proximate care; the regression plots can be found in Supplementary Figure 7. However, mean scores for EQ-5D-3L health utility index were negatively associated (indicating poorer health status), but not significantly, with increasing agitation severity in both proximate caregiving (regression coefficient [95% CI]: –0.004 [–0.012, 0.003]; p = 0.262) and the remote caregiving (regression coefficient [95% CI]: –0.006 [–0.017, 0.004]; p = 0.232). We also observed a similar relationship (not significant) between mean scores of EQ–5D–3L VAS and increasing agitation severity in proximate caregiving (regression coefficient [95% CI]: –0.550 [–1.354, 0.254]; p = 0.178). This relationship was far weaker for remote caregiving (regression coefficient [95% CI]: –0.101 [–1.212, 1.416]; p = 0.876).

Caregiver productivity

WPAI-assessed caregiver presenteeism, overall work impairment, and total activity impairment were positively, but non-significantly, associated with increasing agitation severity score in both remote and proximate caregiving, while significant relationships in regression analyses were observed only in presenteeism and total activity impairment for proximate caregivers (p = 0.019 and p < 0.001, respectively; Fig. 6). A non-significant association for caregiver absenteeism to decrease with increasing agitation severity was observed in both the remote and proximate caregiving was observed (Fig. 6). The regression plots can be found in Supplementary Figure 8.

Fig. 6

Linear regression analysis of WPAI-assessed productivity and activity by agitation score and type of care. Proximate Care, caregiver living with patient; Remote Care, caregiver not living with patient. A regression coefficient significantly different from 0 indicates the predicted outcomes differs significantly for different agitation scores; a coef > 0 implies a worse outcome with higher agitation scores, a coef < 0 implies a better outcome with higher agitation scores. All regressions were controlled for patient demographics (age and sex) and clinical characteristics (time since diagnosis, current MMSE score). Coef, coefficient; CI, confidence interval; MMSE, Mini Mental State Examination; WPAI, Work Productivity and Activity Index.

DISCUSSION

In this large, point-in-time survey of physicians and non-professional caregivers, we observed that the burden of caregiving for non-professional caregivers increased with increasing severity of agitation in patients with dementia. Furthermore, caregiver presenteeism, overall work impairment and total activity impairment were greater with more severe agitation. We also explored how these outcomes were associated with agitation considering only remote, and only proximate, caregiving.

The profile of patients experiencing agitation indicated that they were generally older and at a more advanced, severe stage of their disease than those without symptoms of agitation. Other US observational studies have also reported increasing agitation to be associated with increasing dementia severity [23, 24]. Similarly, patients receiving remote care were older and had more advanced, severe dementia, and more severe agitation symptoms than those receiving no care or proximate care. The relationship between agitation and remote/proximate caregiving may be confounded by the care setting, as more patients receiving remote care than those receiving proximate care lived in a nursing home. The observation that greater dementia severity was associated with a higher proportion of remote caregiving might be expected, as the care needs of advanced dementia can make home care extremely challenging [25, 26].

We observed partners/spouses to be the most common caregivers in both agitated and non-agitated patients. The high proportion of partners/spouses providing care that we observed has been reported in a systematic review of studies of non-professional caregivers for dementia [27]. However, we found that, according to physicians, although the majority of patients receiving proximate care were cared for by their partner/spouse, fewer than 10% of patients received remote care from partners/spouses. Only one caregiver (2%) of those completing record forms and providing remote care was the partner/spouse of the patient.

In all analyses, there was a trend for the time spent caregiving to increase with increasing agitation severity, although this was significant only when considering proximate caregiving (p = 0.001). More caregiving time was reported for proximate than remote caregiving giving at all levels of agitation. The mean hours of caregiving reported for patients with no agitation symptoms and those with an agitation severity score of 6 were 52.9 h and 58.6 h, respectively, for all caregivers, 12.3 h and 14.7 h, respectively, for remote caregivers, and 60.9 h and 81.5 h, respectively, for proximate caregivers. The association of caregiving time with agitation severity and remote/proximate caregiving must be interpreted with care as this will be affected by the patient’s living situation, with a lower likelihood of patients with more severe dementia and agitation symptoms living at home. To the authors’ knowledge, this is the first published analysis exploring the association of agitation symptoms and remote/proximate caregiving on caregiving time for dementia patients.

When considering all patients, all elements of caregiver burden increased with increasing agitation score. A search of the literature failed to identify other US point-in-time studies of the relationship between caregiver burden and patient agitation in dementia; however, a three-year Australian longitudinal study reported higher caregiver burden (measured via the ZBI) for patients with symptoms of agitation than those without [28], and a three-year longitudinal study in Spain found increased ZBI-assessed caregiver burden to be associated with the patient experiencing more severe neuropsychiatric symptoms [29].

Regression analyses of data related to all patients showed that as agitation severity increased, caregiver presenteeism, overall work impairment and total activity impairment worsened. To our knowledge, this is a novel finding, as a literature search identified no studies showing links between agitation severity in dementia patients and caregiver productivity. Further research is needed to confirm this finding and better understand the economic impact of agitation in dementia on caregivers’ work environments.

Analysis of caregiver burden in remote versus proximate caregiving revealed interesting findings, with ZBI Total Caregiver Burden, Personal Strain, and Role Strain all increasing with increasing agitation severity for proximate caregivers and ZBI Guilt increasing with increasing agitation severity for remote caregivers. As many patients receiving remote care were in nursing homes, results suggest that guilt associated with placing a dementia patient in the nursing home setting may be a source of stress for remote caregivers. These results are consistent with published evidence. For example, a Spanish observational study reported increased ZBI Total Caregiver Burden to be associated with the caregiver living with the patient [29], but not remote, caregivers. Furthermore, in a recent study, caregiver burden was found to be higher in caretakers of younger patients, male patients, or parents, which was reflected in our finding that burden was higher in proximate care and patients in proximate care tended to be younger, were more likely to be cared for by their children, and that a lower number of female patients were in proximate care. Lower guilt, however, was associated with older patients who are more likely to be in remote care potentially indicating that the guilt of placing a patient in remote care plays a larger role in caretaker burden than age [5]. Presenteeism and total activity impairment increased with increasing patient agitation severity for proximate caregivers, but not remote caregivers. This finding may also be related to the higher proportion of remote care patients residing in nursing homes, as work-related impairment may have been reduced for caregivers who had the additional support of nursing home staff.

Despite finding that both increased agitation and proximal care setting significantly increased caregiver burden we were unable to find any impact on caregiver HRQoL as measured by the EQ-5D-3L and EQ-VAS. This was perhaps unsurprising as it has been shown that EQ-5D and ZBI scores correlate poorly. EQ-5D showed little difference in scores as disease severity increased and low sensitivity to changes in caregiver circumstance [30]. Our finding that HRQoL did not correlate with increase burden does not necessarily indicate that increased burden does not affect HRQoL, it may be that a generic HRQoL measure such as the EQ-5D fails to capture the unique challenges to HRQoL faced by those caring for AD patients.

A number of limitations should be considered in the evaluation of our findings. Some limitations reflect the survey methodology and are commonly seen with this methodological approach. This was not a true random sample; physicians were asked to provide data for a consecutive series of patients to avoid selection bias, but patients may be those who visit their physician more frequently than the general dementia population. This criterion might result in selecting patients with more severe symptoms, but this bias would be offset to some degree as physicians were allowed to select target patients. Identification of the target patient group was based on the judgement of the respondent physician, and physicians may have been less likely to select severely ill patients. In addition, the survey excluded physicians based in residential care facilities, such as nursing homes; thus, more severe dementia cases might not be fully represented in this sample. It is also worth noting that the caregiver report form was entirely voluntary, this could risk selecting for caregivers with the lowest burden as they have more time to fill out the form, or possibly those with the highest burden as they are more motivated to report on their burden

Recall bias might have affected physician and caregiver responses; however, data were collected at the time of the patient’s appointment, reducing the likelihood of recall bias. The point-in-time design of the survey prevents conclusions about causal relationships, although identification of significant associations is possible. Our methods ensured physicians were unaware of responses on caregiver-reported forms, but it was not possible to confirm no information exchange between physicians and caregivers. Despite data collection taking place across 2015/2016, little has changed in terms of management and impact of disease, therefore, the data are likely suggestive of contemporaneous burden of agitation.

Other limitations are more specific to this analysis. We were not able to evaluate differences in our results attributable to physician specialty, although we recognize that the level of knowledge and management strategies for dementia might differ between primary care physicians and specialists. There is no firm consensus on the definition of agitation; in this survey, the presence of one or more of seven agitation-related physician-reported symptoms was sufficient to consider a patient to be agitated. We acknowledge that other definitions of agitation (such as that used in the NPI-Q) would define agitation differently than the definition used in this study. The definition of agitation used was constructed to best utilize the data available through the DSP and reflects the IPA consensus definition of agitation [7]. In addition, the IPA definition was defined more recently than NPI-Q and reflects a more current understanding of the wide range of agitation symptoms. The method employed in calculating agitation severity scores gave equal and weighting to all symptoms and assumed a linear relationship between mild, moderate, and severe symptom severity. Care must be taken in interpreting the associations observed between agitation severity, remote/proximate caregiving, and hours of caregiving. Given that a higher proportion of patients reside in a nursing home as dementia advances and agitation worsens, remote and proximate care experiences may be qualitatively different for caregivers. Additionally, we chose to run separate analyses on care setting and agitation rather than a covariant analysis to increase data flexibility and ease of interpretation, this did, however, lower the power of our analyses potentially meaning some significant associations appeared not to be significant.

Despite such limitations, real world studies play an important part in highlighting areas of concern that are not addressed in clinical trials. Moreover, there is a paucity of published data on the association of agitation with caregiver burden and productivity in dementia, hence the importance of our analysis.

We have reported separately that increased healthcare resource utilization and direct healthcare costs were associated with increased severity of agitation symptoms in patients with dementia, based on analysis of the data source utilized for the current analysis [31]. The focus of the current survey was on the humanistic impact of agitation in dementia on caregivers, but our findings highlight further potential cost implications of agitation in dementia. The increase in caregiving time and lost productivity of caregivers caring for patients with agitation will certainly translate into higher indirect cost. Lost caregiver productivity needs further study, as an addition to the considerable cost burden of dementia, which is already well documented [32, 33].

In conclusion, our data demonstrated a strong relationship of agitation with burden and productivity impairment for non-professional caregivers. We observed agitation severity to be associated with higher burden and reduced productivity in non-professional caregivers, particularly those providing proximate caregiving.