Patient and Caregiver Benefit From a Comprehensive Dementia Care Program

1-Year Results From the UCLA Alzheimer's and Dementia Care Program

David B. Reuben, MD, AGSF; Zaldy S. Tan, MD, MPH; Tahmineh Romero, MS; Neil S. Wenger, MD, MPH; Emmett Keeler, PhD; Lee A. Jennings, MD, MSHS


J Am Geriatr Soc. 2019;67(7):2267-2273. 

In This Article


From July 1, 2012, to December 2014, 1091 patients and their caregivers entered the program; 991 patients survived for at least 1 year in the program, and 554 patients (56%) and 469 caregivers (47%) caregivers had completed outcome measures at year 1. Loss to follow-up at year 1 included 247 patient-caregiver dyads who were not actively involved in the program due to relocation, change in eligibility (eg, living in a nursing home or enrolled in hospice), or failing to respond to program outreach efforts as well as 190 patients and 275 caregivers who remained in the program but did not complete 1-year surveys. Baseline sociodemographic characteristics of patients and their caregivers with outcome data were similar to the sample missing outcomes but had slightly better baseline scores on the FAQ, Cornell Scale for Depression in Dementia, and both subscales of the NPI-Q that were statistically, although not clinically, significant (Table 2).

After 1 year in the program, patients' cognition (MMSE) and functional status (FAQ) worsened, but behavioral and psychological symptoms (NPI-Q Severity) and depressive symptoms (Cornell) improved (Figure 1A). At 1 year, all caregiver outcomes improved significantly (Figure 1B). In multivariable linear regression models (Table 3), the only predictor of patient change on NPI-Q Severity scale was baseline caregiver NPI-Q Distress score (worse baseline scores predicted better 1-year scores). Predictors of caregiver change scores on the NPI-Q Distress scale included baseline NPI-Q Severity and baseline MCSI (worse baseline scores predicted better 1-year scores). Predictors of change scores on MCSI were black race (predicted better 1-year score vs white race) and functional status measured by FAQ (worse baseline scores predicted better 1-year scores). Predictors of PHQ-9 improvement were child caregiver (predicted better 1-year score vs spouse) and baseline MCSI (worse baseline scores predicted better 1-year scores).

Figure 1.

Baseline and 1-year patient and caregiver outcomes measures. A, Patient outcome measures. B, Caregiver outcome measures. Cornell indicates Cornell Scale for Depression in Dementia; DBS-CG, Dementia Burden Scale-Caregiver; FAQ, Functional Activities Questionnaire; MCSI, Modified Caregiver Strain Index; MMSE, Mini-Mental State Examination; NPI-Q, Neuropsychiatric Inventory Questionnaire; PHQ-9 = Patient Health Questionnaire-9.

Using the above definitions of clinical benefit (Table 1), 314/543 (58%) of patients and 282/447 (63%) of caregivers demonstrated benefit at 1 year. When considering benefit as for the patient, the caregiver, or both, 376/501 (75%) demonstrated benefit at 1 year. Among patients who derived benefit at 1 year, 188/314 (60%) was symptom improvement and 127/314 (40%) was maintaining low symptoms. Among caregivers who benefited at 1 year, 176/282 (62%) was by improving symptoms and 106/282 (38%) was maintaining low symptoms.

Unadjusted bivariate baseline predictors of 1-year clinical benefit are presented in Supplementary Table S1. In adjusted multivariable models (Table 4), at 1 year, patients who had more behavioral symptoms at baseline and fewer depression symptoms were more likely to improve and those with fewer baseline depression symptoms were more likely to maintain low behavioral symptoms. Being a male caregiver, caring for a patient with fewer depressive symptoms, and higher baseline caregiver burden were associated with caregiver improvement. Being a male caregiver and more baseline patient functional impairment, fewer patient depressive symptoms, and fewer patient behavioral symptoms were associated with caregivers who were more likely to maintain low burden at 1 year.

In the sensitivity analyses using unweighted multivariable logistic models (Supplementary Table S2), the magnitude and direction of all the point estimates were similar to IPW models; however, a few of the predictors were no longer statistically significant, plausibly due to reduced power. Specifically, patient depressive symptoms were not statistically significant in predicting patient improvement; and in the model predicting caregiver maintaining low symptoms, caregiver sex, functional impairment, and patient behavioral symptoms were no longer statistically significant.