Results

The sociodemographic characteristics of the caregivers are presented in Table 1. The dementia caregivers had a mean age of 63 years (SD = 11 years), and the majority were female (77%). Most participants had completed at least a high school education (90%), and slightly more than half had a spouse or partner. About half of the participants reported their health as poor or fair. Most caregivers were the adult children of the care recipient (75%) and reported providing an average of around 100 hr of care per week. Table 2 presents the frequency of BPSD along with the severity distribution among those who reported each behavior or symptom. Participants reported that care recipients exhibited an average of five concurrent BPSD symptoms, with nighttime behavior being the most frequently reported symptom (n = 65) and delusions being the least frequent (n = 22).

Variable | Mean | SD | Range | %
Age | 63.0 | 10.9 | 28–87 | 
Female |  |  |  | 77.2
Educational attainment |  |  |  | 
Less than high school |  |  |  | 9.9
High school |  |  |  | 43.6
Associate degree or technical degree |  |  |  | 18.8
Bachelor’s degree or higher |  |  |  | 27.7
Married or partnered |  |  |  | 54.5
Self-rated health |  |  |  | 
Poor or fair |  |  |  | 54.5
Good |  |  |  | 32.7
Very good or excellent |  |  |  | 12.9
Relationship with care-recipient |  |  |  | 
Spouse or partner |  |  |  | 9.9
Son or daughter |  |  |  | 75.2
Others |  |  |  | 14.9
Weekly caregiving hours | 99.4 | 60.1 | 5–168 | 
Less than 35 |  |  |  | 19.8
Between 35 and 100 |  |  |  | 29.7
More than 100 |  |  |  | 50.5
Number of concurrent BPSD | 4.9 | 3.1 | 0–12 | 
Perceived social support | 14.0 | 5.8 | 3–30 | 
Caregiver burden | 7.9 | 6.3 | 0–24 | 
Depressive symptoms | 4.3 | 2.8 | 1–12 | 

Symptoms | Yes | Mild | Moderate | Severe
Nighttime behavior | 65 | 8 | 37 | 20
Apathy/indifference | 54 | 16 | 20 | 18
Agitation/aggression | 53 | 17 | 19 | 17
Appetite/eating | 46 | 14 | 17 | 15
Irritability/lability | 43 | 14 | 16 | 13
Dysphoria/depression | 43 | 22 | 15 | 6
Aberrant motor | 39 | 10 | 20 | 9
Hallucinations | 38 | 12 | 15 | 11
Disinhibition | 37 | 18 | 8 | 11
Anxiety | 31 | 7 | 9 | 15
Euphoria/elation | 26 | 11 | 13 | 2
Delusions | 22 | 7 | 9 | 6

Supplementary Table 1 shows the polychoric correlation ­matrices derived from the 12 NPI-Q items. Exploratory factor analysis identified two BPSD factors, which were generally represented by emotional dysfunction/behavioral instability (eigenvalue = 3.51, variance = 0.29) and perceptual disturbances/apathy (eigenvalue = 2.03, variance = 0.17). Using a minimum loading threshold of 0.4, 7 of the 12 BPSD items loaded onto emotional dysfunction/behavioral instability in the following order of ­loading magnitude: delusions, disinhibition, agitation/aggression, ­irritability/lability, aberrant motor behavior, anxiety, and nighttime behavior. Four BPSD items loaded onto perceptual disturbances/apathy, in the following order: dysphoria/depression, appetite/eating disturbances, apathy/indifference, and hallucinations. Cronbach’s α was 0.80 for emotional dysfunction/behavioral instability and 0.71 for perceptual disturbances/apathy. Among the 12 BPSD items, euphoria/elation had loadings below 0.4 on both factors (Supplementary Table 2). Since it was excluded from the CFA due to low factor loadings, the symptom was also not included in the OLS analyses. CFA of the two-factor model derived from EFA yielded _χ_2(43) = 56.44, CFI = 0.97, TLI = 0.96, and RMSEA = 0.06 (90% CI: 0.00, 0.09), supporting the model’s adequacy. The correlations among all study variables, including the BPSD factors derived from factor analysis, are summarized in Supplementary Table 3.

Table 3 presents the main associations between caregiver burden and the two BPSD factors and the outcome of depressive symptoms. The results indicated that higher caregiver burden and each of the BPSD factors were significantly and positively associated with depressive symptoms across all three models. In addition, the estimates for caregiver burden and the two factors in relation to depressive symptoms remained consistent after the inclusion of weekly caregiving hours in Model 2 and the inclusion of perceived social support in Model 3, suggesting that perceived social support had no significant mitigating effect on the magnitude of these associations.

Variable | Model 1 | Model 2 | Model 3
Est. | SE | β | p | Est. | SE | β | p | Est. | SE | β | p
Intercept | −1.20 | 1.26 | 0.00 | .344 | −1.09 | 1.28 | 0.00 | .397 | −0.21 | 1.34 | 0.00 | .873
Caregiver burden | 0.31 | 0.03 | 0.70 | <.001 | 0.32 | 0.03 | 0.70 | <.001 | 0.30 | 0.03 | 0.68 | <.001
BPSD Factor 1 | 1.45 | 0.25 | 0.51 | <.001 | 1.46 | 0.25 | 0.51 | <.001 | 1.41 | 0.24 | 0.50 | <.001
BPSD Factor 2 | 1.47 | 0.24 | 0.52 | <.001 | 1.49 | 0.25 | 0.53 | <.001 | 1.42 | 0.24 | 0.50 | <.001
Caregiver burden × Factor 1 | 0.05 | 0.03 | 0.18 | .104 | 0.05 | 0.03 | 0.19 | .095 | 0.04 | 0.03 | 0.15 | .169
Caregiver burden × Factor 2 | 0.06 | 0.03 | 0.23 | .035 | 0.06 | 0.03 | 0.24 | .032 | 0.05 | 0.03 | 0.21 | .060

Finally, we examined the moderating effects of each BPSD factor on the association between caregiver burden and depressive symptoms (see also Table 3). Only perceptual disturbances/apathy showed a significant moderating effect in Models 1 and 2 (β = 0.23, p < .05; β = 0.24, p < .05), and a non-significant effect in Model 3 (β = 0.21, p = .060), such that the association between caregiver burden and depressive symptoms became stronger as BPSD symptoms reflected in perceptual disturbances/apathy increased. The association between caregiver burden and depressive symptoms was statistically significant across the entire spectrum of perceptual disturbances/apathy scores (see Figure 2).

Figure 2: Association between caregiver burden and depressive symptoms, with perceptual disturbances/apathy as a moderator.Note. The region of significance from the Johnson-Neyman analysis is highlighted; The graph above corresponds to Model 3 and presents results adjusted for covariates including age, sex, education, marital status, self-rated health, hours of caregiving per week, and perceived social support.

In post hoc analyses, we sought to obtain a better understanding of potential drivers behind the observed moderation effect by conducting follow-up analyses using the individual BPSD ­symptoms with the highest loadings on the perceptual disturbances/apathy factor: dysphoria/depression (loading = 0.73) and appetite/eating (loading = 0.60; Supplementary Table 2). Only appetite/eating consistently showed a statistically significant moderating effect across all three models (Model 1: β = 0.43, p < .01; Model 2: β = 0.42, p < .01; Model 3: β = 0.37, p < .05), indicating that the association between caregiver burden and depressive symptoms became markedly stronger as appetite/eating problems increased (Supplementary Table 4). The standardized coefficient β for the moderating effect of the appetite/eating symptom alone was also nearly twice the size observed for perceptual disturbances/apathy (0.21 vs. 0.37).