Can the Modified Caregiver Strain Index Serve as a Proxy for the Zarit Burden Interview? Insights From an Established GUIDE Participating Organization

Abstract

Receiver Operating Characteristic Curve for the Modified Caregiver Strain Index predicting high Zarit‐Burden Interview, showing excellent discrimination.

1. Introduction

The Centers for Medicare & Medicaid Services (CMS) launched the Guiding an Improved Dementia Experience (GUIDE) Model on July 1, 2024 to deliver comprehensive dementia care, support caregivers, and help people living with dementia (PLWD) remain in their communities [1]. The 22‐item, self‐reported questionnaire Zarit Burden Interview (ZBI) [2] is the required caregiver tool used to categorize patients with moderate to severe dementia into moderate‐ versus high‐dementia complexity tier for the GUIDE Model [3]. However, its length contributes to survey fatigue in the GUIDE Model, which already requires numerous other evaluation measures. The Modified Caregiver Strain Index (MCSI) is a shorter, 13‐item questionnaire that assesses caregiver strain across similar domains [4]. The UCLA Alzheimer's and Dementia Care (ADC) Program, an established GUIDE Model participant, has historically used the MCSI [5, 6]. This study evaluates the relationship between the ZBI and MCSI to determine whether one measure can serve as a proxy for the other.

2. Methods

We collected data from GUIDE Model initial assessments for patients enrolled from July 1, 2024 to June 30, 2025, including patient age and caregiver characteristics (age, sex, race and ethnicity, relationship to the PLWD, living situation, years in the caregiving role, ZBI, and MCSI). Continuous variables were reported as medians with interquartile ranges (IQRs); categorical variables as counts with percentages. Correlations between the ZBI and MCSI were examined using Spearman's ranked correlation coefficient (ρ) and a scatterplot with locally weighted regression (lowess) curve was constructed to visualize the relationship between the two scales (Supplemental Figure S1).

We constructed a logistic regression model to predict high caregiver burden, defined as ZBI score 61–88 based on CMS GUIDE methodology criteria [3]. Predictors included MCSI score, caregiver age, sex, and years in the caregiving role. Predicted probabilities from the model were used to construct a receiver operating characteristic (ROC) curve, and discrimination was evaluated using the area under the curve (AUC) with 95% confidence intervals (CIs). Model calibration was assessed using the Hosmer–Lemeshow goodness‐of‐fit test. Sensitivity and specificity of the MCSI at all possible thresholds for predicting high caregiver burden (ZBI ≥ 61) were examined. Three candidate cut points (≥ 16, ≥ 18, ≥ 20) were selected based on test performance and corresponding sensitivity, specificity, and likelihood ratios were reported.

3. Results

Among 533 caregiver‐patient dyads, the median age was 83 years (IQR 77–87) for PLWD and 64 years (IQR 54–74) for caregivers (Table 1). Caregivers were predominantly female (71%), non‐Hispanic White (55%), children (55%), and in a caregiving role for a median of 5 years (IQR 3–6). About 63% of caregivers lived with the PLWD.

TABLE 1.

Characteristics of caregivers of persons living with dementia enrolled in the GUIDE program at UCLA from July 1, 2024 to June 30, 2025.

Caregiver characteristics	Sample (n = 533)
Caregiver age, median (IQR), years	64 (54–74)
Female sex, n (%)	381 (71%)
Race and ethnicity, n (%)
Non‐Hispanic White	291 (55%)
Black or African American	35 (7%)
Asian	58 (11%)
Hispanic or Latino	48 (9%)
Other a	101 (19%)
Relationship to Patient, No. (%)
Spouse/Partner	181 (34%)
Child	295 (55%)
Other Relationship b	57 (11%)
Years in caregiving role, median (IQR)	5 (3–6)
Lives with patient, n (%)	335 (63%)
High burden by ZBI, n (%)	42 (8%)
Median ZBI total score (IQR)	33 (20–47)
Median MCSI total score (IQR)	11 (6–17)

^a Other = American Indian or Alaska Native, Other Race, Do not identify with Race, Declined to Answer.

^b Other Relationship = Sibling, Other family member, Friend, Non‐family member friend.

Median scores were 33 (IQR 20–47) for the ZBI and 11 (IQR 6–17) for the MCSI; 8% (42/533) of caregivers met criteria for high burden on the ZBI. MCSI was strongly correlated with the ZBI total score (ρ = 0.77, p < 0.001), with higher MCSI scores corresponding with higher ZBI scores (Figure S1). In logistic regression models predicting high caregiver burden (ZBI ≥ 61) and adjusted for caregiver age, sex, and years caregiving, the MCSI showed excellent discrimination (AUC = 0.89; 95% CI, 0.85–0.94; Figure 1) and good calibration (Hosmer–Lemeshow p = 0.78). The ROC curve showed separation from the reference line, indicating reliable classification of caregivers with high versus lower burden. MCSI cut‐point of ≥ 18 provided the best balance of sensitivity (83%) and specificity (82%) (LR+ 4.60; LR− 0.20). A threshold of ≥ 16 yielded greater sensitivity (88%) with lower specificity (76%), whereas ≥ 20 improved specificity (87%) but reduced sensitivity (69%) (Table S1).

FIGURE 1.

ROC curve for MCSI predicting high ZBI. Receiver operating characteristic (ROC) curve for prediction of high caregiver burden (Zarit Burden Interview score ≥ 61). The area under the ROC curve was 0.89 (95% CI, 0.85–0.94). The diagonal reference line represents classification by chance.

4. Discussion

Caregiver burden is a core component of determining care complexity and payment for the GUIDE Model. Consistent with findings in caregivers of patients with Parkinson's disease [7], we found a strong correlation between the MCSI and ZBI, with MCSI also showing excellent discrimination in identifying high caregiver burden. An MCSI cut‐point of ≥ 18 provided a balance of sensitivity and specificity. These findings suggest that the MCSI, a shorter and less burdensome instrument, can serve as a practical substitute for the ZBI in identifying caregivers experiencing high levels of burden.

Although our study was limited to a single GUIDE Model site, it included a relatively large sample of GUIDE caregivers. The MCSI offers a practical, lower‐burden alternative to the ZBI that may help reduce assessment fatigue while preserving accuracy in identifying high caregiver strain within the GUIDE Model.

Author Contributions

All authors contributed to the conceptualization and design of the study. Lee and Centeno were part of the data acquisition and analysis. Lee and Reuben drafted the manuscript. All authors were involved with the interpretation of the data and made critical revisions to the manuscript.

Funding

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Modified Caregiver Strain Index and Zarit Burden Interview total scores correlation. Relationship between Modified Caregiver Strain Index (MCSI) and Zarit Burden Interview (ZBI) scores. Each point represents an individual caregiver. The red line represents a locally weighted regression smoother.

Table S1: jgs70223‐sup‐0001‐supinfo.pdf.