Caregiving: what the evidence shows

Result: benefit (burden, distress, depression)

• varied; in-person (10 studies), remote (2), hybrid (2) in the burden meta-analysis

What was measured: burden, distress, depression

What was found: Post-intervention improvement: caregiver burden MD 8.7 (95% CI 4.1-13.3; I2=89.9%); distress MD 4.3 (0.5-8.2; I2=66.5%); depression MD 5.9 (1.7-10.1; I2=67.0%). Burden by delivery: in-person MD 6.4 (1.4-11.5), remote MD 8.0 (1.9-14.1), hybrid MD 20.9 (5.8-36.0)

Study design: systematic review and meta-analysis of RCTs to July 2025 (17 global and 12 regional databases); RoB 2

Strength: Moderate (24 of 32 RCTs pooled; very high heterogeneity; sensitivity analyses excluding high-risk studies unchanged)

Limitations: Mean differences pooled across different scales; I2 up to 90%; authors could not identify an optimal component combination

Source: Craig et al., BMJ Global Health 2026 · 2026 · PubMed 41506756

Result: benefit (mindfulness, support group); mixed (behaviour management: short-term harm, later benefit)

• 250 people
• components tested over 6 months; 12-month follow-up
• p = <.001 (mindfulness on depression, 12 months)

What was measured: physical-health, burden, perceived-stress, psychological-well-being, anxiety, depression, social-support

What was found: Mindfulness component at 12 months: depressive symptoms beta=-2.13 (95% CI -2.85 to -1.38; P<.001), social support beta=4.76 (1.28-8.15; P=.007). Support-group component: social support beta=4.63 (1.32-7.85; P=.006). Behaviour-management component initially increased anxiety (beta=1.43, 0.43-2.42; P=.005) at 6 months but improved psychological well-being (beta=3.52; P=.008) at 12 months. Mindfulness enhanced the depression benefit of self-care (beta=-1.70; P=.01) and behaviour management (beta=-1.40; P=.04)

Study design: assessor-blinded RCT with 2^5 fractional factorial design (16 conditions); multiphase optimization strategy

Strength: Moderate (factorial RCT n=250; component-level inference; not yet validated as a package)

Limitations: Chinese caregivers in Hong Kong, mean age 48.9; short-term anxiety increase with behaviour-management training; optimized package not yet tested against control

Source: Kwok et al., JAMA Network Open 2025 · 2025 · PubMed 40094667

Result: benefit (outcome-specific by intervention class)

• varied

What was measured: depression, anxiety, quality-of-life, burden

What was found: Depression: ACT, behavioural activation, mindfulness, multicomponent, psychoeducation and CBT 'might reduce' depression. Anxiety: psychoeducation was the only effective intervention. Quality of life: only support groups had a statistically significant effect. Burden: case management, psychoeducation and multicomponent interventions effective (numeric estimates not in abstract)

Study design: systematic review and network meta-analysis of 85 RCTs from seven databases; Cochrane risk-of-bias tool

Strength: Moderate (85 RCTs; ranking evidence, magnitudes not reported in abstract)

Limitations: Abstract reports which interventions were effective without effect sizes; network meta-analysis assumptions (transitivity) across heterogeneous caregiver programmes

Source: Sun, Ji, Leng, Li et al., International Journal of Nursing Studies 2022 · 2022 · PubMed 35247788

Result: benefit (depression); null/weak (burden, anxiety)

• varied

What was measured: depression, quality-of-life, mastery, communication-skills, anxiety, social-support, burden

What was found: Depression most modifiable across psychoeducation, counseling/psychotherapy, occupational therapy, mindfulness and multicomponent programs; evidence for QoL (psychoeducation), mastery and communication skills; null or weak results for anxiety, social support and burden; support groups and respite generally ineffective

Study design: meta-review (umbrella review) appraised with AMSTAR II

Strength: Moderate (most included reviews low quality per AMSTAR II)

Limitations: Review-level evidence; inconsistent intervention classification across reviews

Source: Cheng & Zhang, BMC Geriatrics 2020 · 2020 · PubMed 32293325

Result: benefit

• varied

What was measured: depression, burden, subjective-well-being, physical-health, positive-aspects-of-caregiving

What was found: Post-intervention effects significant for all outcomes when pooled; educational programs with psychotherapeutic components, counseling/psychotherapy and mindfulness had strongest effects on depressive symptoms; multicomponent interventions had largest effects on burden/stress (numeric pooled values not in abstract)

Study design: systematic review and meta-analysis of RCTs (350 post-intervention effect sizes)

Strength: Moderate-high

Limitations: Smaller-sample studies (mindfulness, psychotherapy) reported larger effects; follow-up effects absent for physical health and positive aspects

Source: Cheng, Li, Losada et al., Psychology and Aging 2020 · 2020 · PubMed 31985249

Result: benefit (CBT large, mindfulness moderate, psychoeducation small); null/negligible (emotional support, cognitive rehabilitation, multicomponent)

• varied
• p = 0.013 (CBT); <0.001 (mindfulness); 0.002 (psychoeducation)

What was measured: depression

What was found: CBT (838 participants) SMD -0.905 (95% CI -1.622 to -0.187; p=0.013); mindfulness (186 participants) SMD -0.578 (-0.881 to -0.275; p<0.001); psychoeducation SMD -0.244 (-0.395 to -0.092; p=0.002); emotional support, cognitive rehabilitation and multicomponent interventions: less than small or non-significant

Study design: systematic review and meta-analysis of RCTs published 2007-2017 (MEDLINE, CINAHL, PsycINFO)

Strength: Moderate (31 RCTs; wide CI for CBT estimate)

Limitations: CBT confidence interval very wide (-1.62 to -0.19); search limited to 2007-2017 and three databases; multicomponent null contrasts with REACH II and other trials

Source: Lee et al., Dementia (London) 2020 · 2020 · PubMed 30638044

Result: benefit

• 6 months; 12 individual in-home/telephone sessions and 5 telephone support-group sessions

What was measured: burden, depression, caregiving-frustration, problem-behaviours, hours-on-duty

What was found: Significant decreases in burden, depression, impact of depression on daily life, caregiving frustrations and number of troubling dementia-related behaviours from baseline to 6 months; ~2-hour/day decrease in time on duty approached significance

Study design: pre-post clinical translation (no control group), 24 VA medical centers in 15 states

Strength: Moderate (uncontrolled translation study; consistent with REACH II RCT)

Limitations: No control group; sample size not given in abstract; VA-specific population

Source: Nichols et al., Archives of Internal Medicine 2011 · 2011 · PubMed 21357811

Result: benefit (odds of institutionalization); uncertain (time to institutionalization)

• 9,043 people
• varied
• p = .05 (odds); .004 (best-quality studies)

What was measured: institutionalization-delay

What was found: Odds of institutionalization OR 0.66 (95% CI 0.43-0.99; P=.05); time to institutionalization SMD 1.44 (0.07-2.81; P=.04). Best-quality studies: OR 0.60 (0.43-0.85; P=.004) but time to institutionalization no longer significant (SMD 1.55, -0.35 to 3.45; P=.11). Actively involving caregivers in treatment choices distinguished effective from ineffective programmes

Study design: systematic review and meta-analysis of support programmes with institutionalization outcomes

Strength: Moderate (13 programmes, 9,043 patients; borderline pooled significance)

Limitations: Wide confidence intervals; older literature (pre-2008); heterogeneous programmes

Source: Spijker et al., Journal of the American Geriatrics Society 2008 · 2008 · PubMed 18410323

Result: benefit

• 642 people
• 6 months; 12 in-home and telephone sessions
• p = 0.001 (clinical depression)

What was measured: quality-of-life, depression, burden, institutionalization-delay, social-support

What was found: Clinical depression prevalence 12.6% intervention vs 22.7% control at 6 months; QoL composite improved significantly for Hispanic/Latino (P<0.001), White (P=0.037) and Black spouse caregivers (P=0.003)

Study design: randomized controlled trial, 5 US sites, stratified by race/ethnicity

Strength: High (multi-site RCT, n=642; primary outcome met)

Limitations: Single 6-month follow-up; no significant difference in institutionalization at 6 months; heterogeneous ethnic groups pooled; some ethnic groups excluded

Source: Belle et al., Annals of Internal Medicine 2006 (REACH II RCT) · 2006 · PubMed 17116917

Result: benefit

• 406 people
• 6 sessions individual and family counseling, support group participation, continuous ad hoc telephone counseling; follow-up over 9.5 years of enrolment
• p = 0.025

What was measured: institutionalization-delay, depression, social-support

What was found: Hazard ratio for nursing home placement 0.717 (28.3% reduction in placement rate); model-predicted median time to placement 557 days longer

Study design: randomized controlled trial vs usual care, single site

Strength: High (RCT, long follow-up)

Limitations: Spouse caregivers only; referred volunteer sample; single metropolitan area

Source: Mittelman et al., Neurology 2006 · 2006 · PubMed 17101889

Result: benefit (small; domain-specific)

• varied

What was measured: burden, depression, subjective-well-being, ability-knowledge, care-recipient-symptoms, institutionalization-delay

What was found: Significant but small average effects on burden, depression, subjective well-being, ability/knowledge and care-recipient symptoms (numeric values not in abstract); only multicomponent interventions reduced risk of institutionalization

Study design: meta-analysis

Strength: Moderate-high (largest meta-analysis of its era)

Limitations: Effect magnitudes not reported in abstract; heterogeneity by caregiver gender and publication year

Source: Pinquart & Sorensen, International Psychogeriatrics 2006 · 2006 · PubMed 16686964

Result: benefit (distress, knowledge); null (burden)

• varied

What was measured: psychological-distress, burden, knowledge, patient-mood, institutionalization-delay

What was found: Caregiver psychological distress ES 0.31 (95% CI 0.13-0.50); caregiver knowledge ES 0.51 (0.05-0.98); any main caregiver outcome ES 0.32 (0.15-0.48); patient mood ES 0.68 (0.30-1.06); caregiver burden ES 0.09 (-0.09 to 0.26, not significant)

Study design: meta-analysis of controlled trials

Strength: Moderate (older literature; heterogeneous)

Limitations: Excludes respite care; English-language studies 1985-2001; considerable outcome variability

Source: Brodaty, Green & Koschera, Journal of the American Geriatrics Society 2003 · 2003 · PubMed 12752841

Result: benefit (small)

• varied

What was measured: burden, depression, subjective-well-being, caregiver-satisfaction, ability-knowledge, care-receiver-symptoms

What was found: Combined interventions improved outcomes by 0.14 to 0.41 SD units on average across the six outcomes; effects larger for ability/knowledge than for burden or depression; effects for dementia caregivers smaller than for other groups

Study design: meta-analysis

Strength: Moderate

Limitations: Not dementia-specific (dementia caregivers a subgroup with smaller effects); older literature

Source: Sorensen, Pinquart & Duberstein, The Gerontologist 2002 · 2002 · PubMed 12040138

Result: benefit

• 206 people
• 6 individual/family counseling sessions within 4 months, plus required support group and ad hoc phone counseling (ongoing)
• p = 0.02

What was measured: institutionalization-delay

What was found: Median time to nursing home placement 329 days longer in treatment group; adjusted relative risk of placement 0.65 (95% CI 0.45-0.94)

Study design: randomized controlled trial, single site (New York)

Strength: High (RCT; long follow-up); single site limits generalizability

Limitations: Referred volunteer sample of spouse caregivers only; single site; enrolment over 3.5 years

Source: Mittelman et al., JAMA 1996 · 1996 · PubMed 8940320

Result: null (primary outcomes); small benefit (self-efficacy)

• 2,176 people
• 18 months

What was measured: caregiver-strain, neuropsychiatric-symptoms, caregiver-self-efficacy

What was found: No significant differences on primary outcomes: Modified Caregiver Strain Index LSM 10.7 (health system) vs 10.5 (community) vs 10.6 (usual care); NPI-Q 9.8 vs 9.5 vs 10.1. Caregiver self-efficacy higher for both interventions vs usual care (differences 0.70 [95% CI 0.26-1.14] and 0.85 [0.42-1.29])

Study design: pragmatic randomized clinical trial, 4 US sites, 7:7:1 randomization

Strength: High (large pragmatic RCT) - null result

Limitations: Usual-care arm small (n=144); 62% completed 18 months; enrolment spanned COVID-19

Source: Reuben et al., JAMA 2025 (D-CARE randomized clinical trial) · 2025 · PubMed 39878968

Result: null (patient symptoms, caregiver depression/distress); benefit (ED visits and hospitalizations)

• 201 people
• 24 months (quarterly assessments)
• p = .87 (primary outcome)

What was measured: neuropsychiatric-symptoms, depression, caregiver-distress, emergency-department-use, hospitalization

What was found: Primary NPI-Q severity: between-group difference at 24 months -0.24 (95% CI -2.33 to 1.84), P=.87 for group-by-time. No significant differences in caregiver depression (PHQ-8) or distress. Combined ED visits and hospitalizations 1.06 vs 2.37 events per patient (difference -1.31, 95% CI -1.93 to -0.69; relative risk 0.45, 95% CI 0.31-0.65)

Study design: randomized clinical trial vs usual care, two sites in central Indiana, enrolled March 2019-December 2020

Strength: High (RCT n=201 dyads, 43.3% African American, >=90% assessment completion; primary outcome null)

Limitations: 83 patients died during follow-up; two sites in one state

Source: Sachs et al., JAMA 2025 (IN-PEACE RCT) · 2025 · PubMed 39878993

Result: not yet reported

• 8-year model test beginning July 1, 2024 (established track); new-program track began delivering services July 1, 2025

What was measured: institutionalization-delay, hospitalization, emergency-department-use, burden, quality-of-life, expenditures

Study design: voluntary nationwide payment-and-delivery model test under Section 1115A; formal CMS evaluation pending

Strength: None yet (model in progress; outcomes are stated aims, not results)

Limitations: No evaluation results released; the closest RCT evidence for similar care-management models is mixed (Care Ecosystem positive, D-CARE null)

Source: CMS Innovation Center GUIDE Model (no outcome results published as of Sept 2026) · 2024

Result: benefit

• 780 people
• 12 months
• p = 0.04 (primary QoL); 0.03 (depression); 0.046 (burden)

What was measured: quality-of-life, emergency-department-use, depression, burden

What was found: PWD quality of life B 0.53 (95% CI 0.25-1.30, P=0.04); ED visits B -0.14 (95% CI -0.29 to -0.01, P=0.04); caregiver depression (PHQ-9) B -1.14 (95% CI -2.15 to -0.13, P=0.03); caregiver burden (ZBI-12) B -1.90 (95% CI -3.89 to -0.08, P=0.046)

Study design: single-blind pragmatic randomized clinical trial, California/Nebraska/Iowa, 2:1 randomization vs usual care

Strength: High (large pragmatic RCT)

Limitations: Caregiver-rated primary outcome; 571 of 780 completed 12-month survey; small absolute effects

Source: Possin et al., JAMA Internal Medicine 2019 (Care Ecosystem RCT) · 2019 · PubMed 31566651

Result: benefit

• 554 people
• 1 year (observational)

What was measured: caregiver-strain, depression, caregiver-distress, burden, neuropsychiatric-symptoms

What was found: 63% of caregivers (282/447) demonstrated clinical benefit at 1 year (improved or maintained low symptoms by MCID); caregivers improved on all scales (Modified Caregiver Strain Index, PHQ-9, NPI-Q distress, Dementia Burden Scale)

Study design: observational pre-post, single academic medical center

Strength: Low-moderate (no control group)

Limitations: Observational; regression to mean possible; single urban academic center

Source: Reuben et al., Journal of the American Geriatrics Society 2019 · 2019 · PubMed 31355423

Result: benefit (small for burden)

• 634 people
• 6 months of home-based care management; 12-month follow-up
• p = 0.045 (caregiver burden)

What was measured: burden, neuropsychiatric-symptoms, quality-of-life, antidementia-drug-treatment

What was found: Caregiver burden b = -0.50 (95% CI -1.09 to 0.08, P=0.045); BPSD b = -7.45 (95% CI -11.08 to -3.81, P<0.001); patient QoL improved only for those not living alone

Study design: pragmatic GP-based cluster-randomized trial, Germany

Strength: Moderate-high (cluster RCT; but 407 of 634 available for primary outcome)

Limitations: Burden CI crosses zero despite reported P=0.045; substantial attrition; German health system

Source: Thyrian et al., JAMA Psychiatry 2017 (DelpHi-MV cluster RCT) · 2017 · PubMed 28746708

Result: null (primary); possible benefit (caregiver time)

• 289 people
• 18 months

What was measured: unmet-needs, burden, quality-of-life, depression, hours-of-care

What was found: No statistically significant between-group differences in unmet caregiver needs, burden, depression or QoL after multiple-comparison adjustment; potentially clinically relevant reduction in caregiver hours spent with care recipient

Study design: 18-month randomized controlled trial (dyads)

Strength: Moderate (RCT) - null caregiver findings

Limitations: Both groups received written needs-assessment results, diluting contrast; single city

Source: Tanner et al., American Journal of Geriatric Psychiatry 2015 · 2015 · PubMed 25260557

Result: benefit (small, time-point dependent)

• 9,615 people
• varied
• p = 0.02 (institutionalization 6 mo); 0.03 (burden 6 mo)

What was measured: institutionalization-delay, hospitalization, burden, depression, quality-of-life, cost

What was found: Institutionalization OR 0.82 (95% CI 0.69-0.98) at 6 months and OR 0.25 (0.10-0.61) at 18 months, but uncertain at 10-12 months (OR 0.95) and 24 months; carer burden SMD -0.07 (-0.12 to -0.01) at 6 months; carer depression SMD -0.08 (-0.16 to -0.01) at 18 months; total service cost SMD -0.07 (-0.12 to -0.02) at 12 months

Study design: Cochrane systematic review and meta-analysis of RCTs

Strength: Moderate (low-moderate risk of bias; considerable heterogeneity)

Limitations: 69% of studies at high risk of performance bias; effects inconsistent across time points; searches to Dec 2013

Source: Reilly et al., Cochrane Database of Systematic Reviews 2015 (CD008345) · 2015 · PubMed 25560977

Result: benefit (time at home, self-rated QoL); null (other outcomes)

• 303 people
• 18 months; delivered by non-clinical community workers with interdisciplinary team oversight

What was measured: institutionalization-delay, unmet-needs, quality-of-life, depression, neuropsychiatric-symptoms

What was found: Adjusted hazard of leaving home (all-cause transition) reduced 37%: HR 0.63 (95% CI 0.42-0.94); significant improvement in self-reported QoL; no group difference in total unmet needs, proxy QoL, NPS or depression

Study design: 18-month randomized controlled trial, Baltimore, MD

Strength: Moderate (pilot RCT, single city)

Limitations: Pilot; single urban area; no significant reduction in total unmet needs

Source: Samus et al., American Journal of Geriatric Psychiatry 2014 (MIND pilot RCT) · 2014 · PubMed 24502822

Result: benefit

• 486 people
• outcomes at 6 and 12 months; telephone and e-mail delivery, caseloads 100-125

What was measured: unmet-needs, caregiver-strain, depression, social-support

What was found: Significant improvements in unmet needs, three types of caregiver strain, depression and two support resources; most improvements evident at 6 months (the abstract does not give their size)

Study design: controlled trial (intervention vs comparison VA sites); 6- and 12-month follow-up

Strength: Moderate (non-randomized comparison-site design)

Limitations: Not individually randomized; 394 and 324 caregivers completed 6- and 12-month follow-up; VA-specific

Source: Bass et al., Journal of the American Geriatrics Society 2013 · 2013 · PubMed 23869899

Result: benefit (caregiver mental health, distress from behaviours); null/trend (burden, behaviours, mortality)

• 81 people
• 6 months (wait-list received intervention after 6 months)

What was measured: psychological-distress, burden, caregiver-distress, neuropsychiatric-symptoms, patient-functional-dependence, mortality

What was found: GHQ caregiver mental health -1.12 (95% CI -2.07 to -0.17); NPI-D caregiver distress -1.96 (-3.51 to -0.41); non-significant reductions in Zarit burden, EASI and NPI-S; non-significant reduction in deaths of people with dementia (OR 0.34, 95% CI 0.01-1.03)

Study design: RCT, immediate intervention vs 6-month wait-list; masked outcome evaluation; mixed-effects models

Strength: Low (small RCT, n=81; 18 deaths during trial)

Limitations: 59 of 81 completed; 18 care recipients died; mild-moderate dementia only

Source: Dias et al., PLoS One 2008 (Goa RCT) · 2008 · PubMed 18523642

Result: benefit (symptoms, caregiver distress/depression); null (placement)

• 153 people
• 12 months of care management by advanced practice nurse; outcomes to 18 months
• p = 0.01 (NPI)

What was measured: neuropsychiatric-symptoms, caregiver-distress, depression, institutionalization-delay, hospitalization

What was found: Total NPI mean difference -5.6 at 12 months (P=0.01) and -5.4 at 18 months (P=0.01); caregiver NPI distress improved at 12 months; caregiver PHQ-9 depression improved at 18 months; no differences in nursing home placement, hospitalization or death

Study design: controlled clinical trial randomized by physician

Strength: Moderate-high (RCT, small)

Limitations: N=153; two university-affiliated systems; caregiver outcomes secondary

Source: Callahan et al., JAMA 2006 · 2006 · PubMed 16684985

Result: benefit (care quality, caregiver support); null (caregiver HRQoL)

• 408 people
• >12 months, care managers in 9 intervention clinics
• p = <0.001 (adherence)

What was measured: guideline-adherence, quality-of-life, caregiving-quality, social-support, unmet-needs

What was found: Guideline adherence 63.9% vs 32.9% (adjusted difference 30.1%, 95% CI 25.2-34.9); caregiving quality, social support and unmet caregiving needs better in intervention (P<0.05); caregiver health-related QoL did not differ

Study design: clinic-level cluster randomized controlled trial, 18 primary care clinics

Strength: High (cluster RCT)

Limitations: Well-educated, predominantly White participants with usual source of care; fee-for-service cost concerns

Source: Vickrey et al., Annals of Internal Medicine 2006 · 2006 · PubMed 17116916

Result: benefit

• 261 people
• 7 weeks; 6-month follow-up

What was measured: positive-dyadic-interaction, dyadic-strain

What was found: Tele-Savvy group reported significant improvements in positive interactions and reduction in dyadic strain over time vs both control groups (the abstract does not give their size)

Study design: secondary analysis of 3-arm RCT (active n=96, attention control n=111, waitlist n=54)

Strength: Low-moderate (secondary analysis)

Limitations: Care-partner-reported only; secondary outcome

Source: Bonds Johnson et al., The Gerontologist 2026 (secondary analysis) · 2026 · PubMed 41734283

Result: small benefit (insomnia); null (sleep quality)

• 137 people
• 7-week virtual programme; 6-month follow-up
• p = 0.050 (ISI)

What was measured: sleep-quality, insomnia

What was found: No initial group-by-time effects for PSQI or ISI; combined Tele-Savvy group showed significant improvement in ISI over time (P=0.050); effect sizes PSQI d=0.126, ISI d=0.310

Study design: secondary analysis of 3-arm RCT

Strength: Low (secondary analysis; borderline P)

Limitations: Not designed for sleep; sample predominantly White (68.6%) and female

Source: Brewster et al., Clinical Interventions in Aging 2023 (secondary analysis of Tele-Savvy RCT) · 2023 · PubMed 38111602

Result: benefit (depression, stress, mastery); null (burden, anxiety)

• 261 people
• 43 days: 7 weekly synchronous group video sessions plus 36 brief asynchronous video lessons

What was measured: depression, perceived-stress, reaction-to-behaviours, caregiver-mastery, burden, anxiety

What was found: Statistically and clinically significant benefits with moderate to large effect sizes for depression, perceived stress, reaction to care-recipient behaviours and mastery (numeric values not in abstract); no benefit for burden or anxiety

Study design: three-arm randomized trial (active, attention control, usual-care waitlist), 2:2:1

Strength: Moderate-high (RCT with attention control; truncated by COVID-19)

Limitations: Trial truncated because of COVID-19 confounding; secondary analysis found benefit mainly in White, not Black, caregivers (PMID 37067896)

Source: Hepburn et al., The Gerontologist 2022 (Tele-Savvy RCT) · 2022 · PubMed 33640979

Result: benefit (affective symptoms); null (care home admission, costs)

• 260 people
• 8 sessions; outcomes at 72 months

What was measured: depression, anxiety, institutionalization-delay, cost

What was found: HADS-T adjusted mean difference -2.00 points (95% CI -3.38 to -0.63) over 72 months; time to care home intensity ratio 0.88 (95% CI 0.58-1.35, not significant)

Study design: randomised controlled trial, 6-year follow-up

Strength: High for durability of mental-health effect

Limitations: 222 of 260 carers analysed at 6 years; care-home and cost differences not statistically significant

Source: Livingston et al., British Journal of Psychiatry 2020 · 2020 · PubMed 31298169

Result: benefit (burden, depression, confidence); null (care-recipient BPSD)

• 60 people
• 6 weekly 2-hour group classes; 6-week follow-up

What was measured: burden, neuropsychiatric-symptoms, depression, self-efficacy, self-rated-health, life-satisfaction

What was found: Caregiver burden d=-0.48; depressive symptoms d=-0.53; self-confidence d=0.68; no significant benefit for care-recipient behavioural and psychological symptoms of dementia. 94% completed at least 4 of 6 classes

Study design: pragmatic two-arm RCT vs usual care (wait-list then PTC); intent-to-treat

Strength: Low (small pragmatic RCT, n=60; 44 at follow-up)

Limitations: Small sample; usual-care arm crossed over after the trial; p-values not in abstract

Source: Terracciano et al., JAMDA 2020 · 2020 · PubMed 31866419

Result: benefit

• 260 people
• 8 sessions delivered by supervised psychology graduates
• p = 0.02 (HADS-T)

What was measured: depression, anxiety, quality-of-life, abusive-behaviour

What was found: HADS total adjusted mean difference -1.80 points (95% CI -3.29 to -0.31) over 8 months; case-level depression OR 0.24 (95% CI 0.07-0.76); carer QoL (HSQ mental health) +4.09 (95% CI 0.34-7.83)

Study design: pragmatic randomised controlled trial, 2:1 randomisation vs treatment as usual, UK

Strength: High (pragmatic RCT; effects sustained at 6 years; cost-effective)

Limitations: Self-identified carers from UK memory services; unblinded participants; no effect on care recipient QoL

Source: Livingston et al., BMJ 2013 (START pragmatic RCT) · 2013 · PubMed 24162942

Result: benefit (psychological morbidity); null (burden, institutionalization)

• varied

What was measured: anxiety, depression, subjective-stress, coping, burden, institutionalization-delay

What was found: Anxiety SMD -0.21 (95% CI -0.39 to -0.04); depression SMD -0.66 (-1.27 to -0.05); subjective stress SMD -0.23 (-0.43 to -0.04); no effects on coping, burden appraisal, reactions to behaviours or institutionalization

Study design: Cochrane systematic review and meta-analysis of RCTs

Strength: Moderate

Limitations: Number of included trials/participants not stated in abstract; search date 5 April 2009; identifying relevant interventions was difficult

Source: Vernooij-Dassen et al., Cochrane Database of Systematic Reviews 2011 (CD005318.pub2) · 2011 · PubMed 22071821

Result: benefit (strain); null (distress, QoL)

• 58 people
• 5 sessions (1 assessment, 2 education, 2 behaviour training); 6-month follow-up

What was measured: caregiver-strain, psychological-distress, quality-of-life, neuropsychiatric-symptoms

What was found: Significantly decreased caregiver strain (Zarit Burden Interview) at 6 months vs controls; no group differences in caregiver psychological distress (SRQ-20) or in caregiver or patient quality of life (values not in abstract)

Study design: RCT, immediate intervention vs 6-month wait-list

Strength: Low (small RCT, n=58)

Limitations: Small; wait-list control; the abstract does not give effect sizes

Source: Guerra et al., Revista Brasileira de Psiquiatria 2011 (10/66 Peru RCT) · 2011 · PubMed 20602013

Result: benefit (burden); null (distress, QoL)

• 60 people
• 5 weekly 30-minute sessions; 6-month follow-up

What was measured: burden, psychological-distress, quality-of-life, neuropsychiatric-symptoms

What was found: Large, statistically significant net improvement in Zarit burden at 6 months vs controls; no group differences in caregiver psychological distress (SRQ-20), caregiver QoL (WHOQOL-BREF) or patient QoL (DEMQOL) (values not in abstract)

Study design: single-blind parallel-group RCT vs medical care as usual

Strength: Low (small RCT, n=60; single outcome significant)

Limitations: Small sample; the abstract does not give effect sizes; one site

Source: Gavrilova et al., International Journal of Geriatric Psychiatry 2009 (10/66 Russia RCT) · 2009 · PubMed 18814197

Result: benefit

• 52 people
• 6 sessions

What was measured: caregiver-mastery, distress

What was found: Experimental participants' mastery and distress scores significantly better than control at 5-6 month follow-up (magnitudes not given in abstract)

Study design: randomized trial with wait-list control, agencies in 3 states

Strength: Low-moderate (small trial; 52 completers)

Limitations: Small completer sample; wait-list design

Source: Hepburn et al., Journal of Gerontological Nursing 2007 · 2007 · PubMed 17378189

Result: benefit at 4 months; no patient group differences at 9 months

• 237 people
• Up to 12 home or telephone contacts over 4 months (OT and nurse)
• p = 0.002 (caregiver well-being)

What was measured: patient-functional-dependence, patient-engagement, caregiver-well-being, caregiver-confidence, quality-of-life

What was found: Caregiver well-being (Perceived Change Index) adjusted mean difference 0.22 (95% CI 0.08-0.36), Cohen d 0.30; caregiver confidence using activities +0.81 (95% CI 0.30-1.32), d 0.54; patient functional dependence d 0.21; IADL dependence d 0.43

Study design: prospective 2-group randomized trial, Pennsylvania

Strength: High (RCT, JAMA)

Limitations: Single state; effects on patients not sustained at 9 months; control received 3 phone calls and educational materials

Source: Gitlin et al., JAMA 2010 (COPE randomized trial) · 2010 · PubMed 20810376

Result: benefit

• 272 people
• Up to 11 home and telephone contacts over 16 weeks, plus 3 reinforcement calls to 24 weeks
• p = 0.002 (primary behaviour outcome)

What was measured: problem-behaviours, caregiver-upset, caregiver-confidence, burden, depression, well-being

What was found: 67.5% of intervention caregivers reported improvement in targeted problem behaviour vs 45.8% of controls (P=0.002); fewer intervention caregivers with depressive symptoms (53.0% vs 67.8%, P=0.02); less burden (P=0.05); better well-being (P=0.001)

Study design: two-group randomized trial vs no-treatment control

Strength: High (RCT, n=272)

Limitations: No-treatment control (no attention control); single region

Source: Gitlin et al., Journal of the American Geriatrics Society 2010 · 2010 · PubMed 20662955

Result: benefit

• 60 people
• 8 occupational therapy sessions over 4 months

What was measured: problem-behaviours, activity-engagement, burden, caregiver-mastery, self-efficacy, hours-on-duty

What was found: Reduced frequency of problem behaviours (especially shadowing and repetitive questioning); fewer caregiver hours doing things/being on duty; greater mastery and self-efficacy (magnitudes not given in abstract)

Study design: randomized controlled pilot study with wait-list control

Strength: Low-moderate (pilot, n=60)

Limitations: Small pilot; wait-list design; single site

Source: Gitlin et al., American Journal of Geriatric Psychiatry 2008 (pilot RCT) · 2008 · PubMed 18310553

Result: benefit

• 95 people
• not stated in abstract (assessments at baseline, post-treatment and 6 months)

What was measured: depression, burden, reactivity-to-behaviours, care-recipient-behaviour-problems, quality-of-life

What was found: Significant improvements in caregiver depression, burden and reactivity to behaviour problems; reduced frequency/severity of care-recipient behaviour problems and improved QoL; maintained at 6 months (the abstract does not give their size)

Study design: randomized controlled trial

Strength: Moderate (small RCT)

Limitations: N=95 dyads; single region; session schedule not given in abstract

Source: Teri et al., The Gerontologist 2005 · 2005 · PubMed 16326662

Result: benefit (patient outcomes); trend (institutionalization)

• 153 people
• 3 months, in-home; follow-up to 24 months
• p = <0.001 (physical role functioning)

What was measured: patient-physical-function, patient-depression, institutionalization-delay

What was found: Patients exercising >=60 min/wk OR 2.82 (95% CI 1.25-6.39); SF-36 physical role functioning mean difference 19.29 (95% CI 8.75-29.83) at 3 months and 10.89 (3.62-18.16) at 2 years; Cornell depression mean difference -1.03; trend toward less institutionalization for behavioural disturbance (19% vs 50%)

Study design: randomized controlled trial vs routine medical care

Strength: High (RCT)

Limitations: Caregiver outcomes not primary; institutionalization difference only a trend; single site (Seattle)

Source: Teri et al., JAMA 2003 · 2003 · PubMed 14559955

Result: benefit (caregiver depression at 12 months); trend (attendance, 6-month depression)

• 203 people
• 12 months
• p = .013 (total depression, 12 months)

What was measured: depression, well-being, adult-day-service-attendance

What was found: 6 months: 43.6% control vs 34.2% ADS Plus caregivers with CES-D >=16 (OR 0.38, p=.072). 12 months: ADS Plus reduced total depression (p=.013) and depressed affect (p=.015) after covariate adjustment. Client attendance 126.05 vs 78.49 days at 9 intervention vs 9 control sites (p=.079; +60.6%)

Study design: cluster-randomized trial in 34 adult day service sites (16 ADS Plus, 18 ADS-only controls)

Strength: Moderate (multisite cluster RCT; 12-month effect significant, 6-month not)

Limitations: 22.7% lost to follow-up at 12 months; caregivers 76.4% White and 88.2% college-educated; attendance data from only 18 of 34 sites

Source: Gitlin et al., The Gerontologist 2024 (ADS Plus multisite cluster-randomized trial) · 2024 · PubMed 37549428

Result: mixed (benefit for burden; possible harm for time to institutionalization)

• varied

What was measured: burden, problem-behaviours, institutionalization-delay

What was found: Day care services effective in decreasing caregiver burden and behavioural problems but accelerate time to nursing home admission; temporary residential admission results mixed with unexpected adverse effects; high-quality evidence on community-based/in-home respite lacking

Study design: systematic review (RCTs, quasi-experimental, pre-post and cohort studies)

Strength: Low-moderate (non-randomized designs included)

Limitations: Few comparable studies; methodological quality variable

Source: Vandepitte et al., International Journal of Geriatric Psychiatry 2016 · 2016 · PubMed 27245986

Result: null / insufficient evidence

• 753 people
• varied

What was measured: burden, psychological-health, institutionalization-delay

What was found: No significant effects of respite care vs no respite on any caregiver variable; vs polarity therapy, caregiver perceived stress favoured polarity therapy (MD 5.80, 95% CI 1.43-10.17, n=38); no evaluable data on people with dementia

Study design: Cochrane systematic review of RCTs (4 trials)

Strength: Very low (Cochrane GRADE rating)

Limitations: Only 4 heterogeneous trials; pooling impossible; authors caution results may reflect lack of high-quality research rather than lack of benefit

Source: Maayan, Soares-Weiser & Lee, Cochrane Database of Systematic Reviews 2014 (CD004396) · 2014 · PubMed 24435941

Result: benefit (day-level)

• 121 people
• Daily assessments before first ADS use and after 1 and 2 months of attendance

What was measured: stressor-exposure, stress-appraisal, behaviour-problems, care-recipient-sleep

What was found: Total exposure to stressors and stress appraisals decreased significantly on ADS days vs non-ADS days; reductions in evening behaviour problems and improved care-recipient sleep after ADS days (the abstract does not give their size)

Study design: within-person withdrawal design (A-B-A-B), no randomized control

Strength: Low-moderate (within-person design; no control group)

Limitations: Non-randomized; caregivers self-selected into ADS; short follow-up

Source: Zarit et al., Journals of Gerontology Series B 2011 · 2011 · PubMed 21642593

Result: benefit

• ongoing programme services (Older Americans Act Title III-E)

What was measured: burden, ability-to-continue-caregiving

What was found: Caregivers receiving 4 or more hours of respite per week had a decrease in self-reported burden over time while comparison caregivers experienced an increase; more respite hours per week increased probability of reporting that services helped them continue caregiving (no numeric effect on loaded page)

Study design: programme evaluation with comparison group (design details not on the loaded page)

Strength: Low (secondary summary of a non-randomized evaluation; primary report not loaded)

Limitations: Not dementia-specific; primary evaluation report not retrieved; numeric values absent on loaded page

Source: NFCSP outcome evaluation as summarized by ARCH National Respite Network ('Respite Facts and Talking Points')

Result: benefit (small to moderate)

• varied

What was measured: psychological-well-being, depression, burden, social-outcomes

What was found: Psychological well-being Hedges g -0.44 (95% CI -0.73 to -0.15); depression g -0.40 (-0.72 to -0.08); burden g -0.23 (-0.33 to -0.13); social outcomes g 0.40 (0.09-0.71)

Study design: meta-analysis of true and quasi-experimental controlled trials (1998-2009)

Strength: Moderate (includes quasi-experimental designs; 'support group' includes psychoeducational groups)

Limitations: Mixes mutual-support with psychoeducational groups; quasi-experimental studies included; contrasts with Cheng 2020 meta-review finding support groups generally ineffective

Source: Chien et al., International Journal of Geriatric Psychiatry 2011 · 2011 · PubMed 21308785

Result: null (burden, mood); small benefit (one QoL measure)

• 93 people
• Online, follow-up at 6 and 12 weeks
• p = 0.003 (CarerQoL-VAS only)

What was measured: burden, depression, quality-of-life, work-productivity

What was found: No significant differences on Zarit Burden Interview, CES-D, WHOQOL-Bref or WPAI:CG; CarerQoL-VAS higher in intervention (mean difference 0.75, 95% CI 0.25-1.24, P=0.003)

Study design: single-blind parallel-group randomised clinical trial vs waitlist, multilingual carers

Strength: Low-moderate (under-recruited: 93 analysed vs 194 target)

Limitations: Recruitment shortfall; possible ceiling effects; short follow-up

Source: Brijnath et al., Age and Ageing 2026 (Draw-Care RCT, Australia) · 2026 · PubMed 41719424

Result: inconclusive (pilot)

• 84 people
• 6 months (manual, online peer meetings, WhatsApp, one-to-one facilitator support)

What was measured: behavioural-problems, carer-distress, quality-of-life, social-support, self-efficacy

What was found: Possible trend toward lower behavioural problems and lower carer distress responses; no clear evidence of effect on carer QoL, social support, self-efficacy or PLWD QoL

Study design: multi-centre internal block-pilot randomised controlled trial with embedded qualitative study

Strength: Very low (pilot; 10% recruitment rate, 70% retention)

Limitations: Not powered for effectiveness; informs main trial only

Source: Yu et al., Nursing Open 2026 (internal pilot of multi-centre RCT) · 2026 · PubMed 42634407

Result: null (primary outcomes at 3 months); benefit (secondary outcomes, distress trajectory)

• 262 people
• 3 months (main endpoint) vs 3-month delayed access
• p = 0.012 (distress time-by-intervention interaction)

What was measured: caregiver-distress, caregiver-confidence, behaviour-problems, patient-functional-dependence, caregiver-upset, well-being

What was found: No treatment differences at 3 months on the three primary outcomes (behavioural symptoms, caregiver distress, confidence); time-by-intervention interaction for distress significant (Bonferroni-corrected p=0.012). Secondary: immediate users reported fewer behaviours (p=0.02), less functional dependence (p=0.05), less associated upset (p=0.005) and better well-being (p<0.001)

Study design: two-group randomized trial, immediate vs 3-month delayed app access; masked interviewers

Strength: Moderate (RCT n=262; primary endpoints not met)

Limitations: Sample 83% White, 93% some college; 3-month endpoint only; the abstract does not give effect sizes

Source: Gitlin, McClure, Webster, Lyketsos & Kales, Journal of the American Geriatrics Society 2026 · 2026 · PubMed 42561014

Result: null (well-being); possible benefit (costs)

• 352 people
• Self-guided online modules; outcomes at 6 months
• p = 0.67 (EQ-5D-5L); 0.003 (cost change)

What was measured: distress, quality-of-life, health-and-social-care-costs, engagement

What was found: Main trial reported no measurable benefits to carer well-being (per process evaluation); EQ-5D-5L no difference (P=0.67); mean change in public-sector resource-use costs favoured intervention (difference GBP 146 per participant, P=0.003)

Study design: multi-centre single-blind randomised controlled trial vs usual care (England, Wales, Scotland)

Strength: Moderate (RCT) - null primary result; primary-outcome paper not we hold

Limitations: 8.3% never used the platform and 54% spent 30-90 minutes total; primary clinical-outcome paper not retrieved in this pass

Source: iSupport UK RCT: cost-consequence analysis (Anthony et al., BMJ Open 2025) and process evaluation (Masterson-Algar et al., Behavioral Sciences 2025) · 2025 · PubMed 40379340

Result: benefit (carer QoL, self-efficacy, social support; care-recipient hospital admissions)

• 149 people
• 12 months
• p = <0.001 (mental-health QoL)

What was measured: quality-of-life, self-efficacy, social-support, reactivity-to-behaviours, hospitalization, emergency-department-use

What was found: At 12 months vs usual care: carer mental-health-related QoL +12.0 points (p<0.001); self-efficacy +14.8 to +18.5 points (p<0.001); social support +0.25 (p<0.028); reactions to behaviour -0.25 (p<0.028); 60% lower hospital admission rate for people with dementia (p=0.045)

Study design: pragmatic randomized controlled trial vs usual care, July 2022-December 2024

Strength: Moderate (pragmatic RCT, n=149; multiple significant outcomes)

Limitations: Small trial; hospital-admission finding borderline (p=0.045); QoL of people with dementia (co-primary) not reported as significant in abstract

Source: Xiao et al., Alzheimer's & Dementia 2025 (pragmatic RCT) · 2025 · PubMed 41023309

Result: null (burden, QoL, social support); benefit (learning behaviours only)

• 120 people
• not stated in abstract (baseline to post-intervention)
• p = .32 (burden, primary)

What was measured: burden, quality-of-life, social-support, learning-behaviours

What was found: No significant between-group differences in caregiver burden (beta=-3.80, 95% CI -11.25 to 3.65; P=.32), quality of life (beta=-1.55; P=.55) or social support (beta=0.85; P=.65); learning behaviours improved (beta=2.35, 95% CI 1.03-3.67; P<.001)

Study design: cluster randomized controlled trial (family-physician-team-guided iSupport + usual care vs usual care + iSupport registration only); intention-to-treat; CFIR implementation analysis

Strength: Low (small cluster RCT, n=120; primary outcome null)

Limitations: Control arm also had iSupport access (registration), diluting contrast; implementation barriers included digital gaps for older users and lack of primary-care incentives

Source: Yuan et al., Journal of Medical Internet Research 2025 · 2025 · PubMed 40925598

Result: benefit (burden, depression, stress; small); null (anxiety)

• 2,132 people
• varied
• p = <0.001 (depression); 0.02 (burden); 0.01 (stress); 0.12 (anxiety)

What was measured: burden, depression, stress, anxiety

What was found: Burden SMD -0.14 (95% CI -0.25 to -0.02; p=0.02); depression SMD -0.17 (-0.27 to -0.07; p<0.001); stress SMD -0.20 (-0.37 to -0.04; p=0.01); anxiety SMD -0.12 (-0.27 to 0.03; p=0.12, not significant). Tailored interventions: depression SMD -0.26 (-0.40 to -0.13) vs standardized SMD -0.08 (-0.22 to 0.06; p=0.25). Internet-based SMD -0.17 and telephone-based SMD -0.18 both effective for depression

Study design: systematic review and meta-analysis of RCTs from nine databases (incl. Chinese) to 26 July 2023

Strength: Moderate (22 RCTs, n=2,132; consistent small effects)

Limitations: Effects small; tailored vs standardized subgroup difference is observational across trials

Source: Zhu et al., Aging & Mental Health 2024 · 2024 · PubMed 38946249

Result: benefit (carer mental-health QoL, self-efficacy, distress reactions)

• 266 people
• 6 months; outcomes at 6 (T1) and 9 months (T2)
• p = 0.002 (mental-health QoL)

What was measured: quality-of-life, self-efficacy, social-support, distress, care-recipient-behaviour-problems, hospitalization, emergency-department-use, institutionalization-delay

What was found: At 6 months vs usual care: mental-health-related QoL mean difference 4.1 (95% CI 1.5-6.8; p=0.002); self-efficacy in controlling upsetting thoughts MD 7.1 (2.2-12.0; p=0.005); distress reactions to changed behaviours MD -0.1 (-0.3 to -0.03; p=0.012)

Study design: multicentre RCT vs usual care; multivariate mixed-effects linear regression; recruited 1 Nov 2021-30 Jun 2022

Strength: Moderate (multicentre RCT, n=266; primary outcome met)

Limitations: Care-recipient outcomes (behaviour frequency, QoL, admissions, nursing-home placement) not reported as significant in abstract; Chinese-language carers only

Source: Xiao et al., International Journal of Nursing Studies 2024 (multicentre RCT) · 2024 · PubMed 39163682

Result: null vs usual care; slight benefit vs information-only (depression)

• 2,367 people
• varied (printed materials, telephone, internet; no face-to-face professional contact)

What was measured: burden, depression, quality-of-life, dropout

What was found: Vs usual care/waitlist/attention control: burden SMD -0.06 (95% CI -0.35 to 0.23; 9 studies); depression SMD -0.05 (-0.22 to 0.12; 8 studies); HRQoL SMD 0.10 (-0.13 to 0.32). Vs information alone: burden SMD -0.24 (-0.51 to 0.04); depression SMD -0.25 (-0.43 to -0.06; 11 studies, 1100 participants); dropout RR 1.51 (1.04-2.20)

Study design: Cochrane systematic review and meta-analysis of RCTs

Strength: Moderate to low certainty (GRADE)

Limitations: Self-rated outcomes by unblinded participants; attrition; remote delivery increased dropout vs information-only

Source: Gonzalez-Fraile et al., Cochrane Database of Systematic Reviews 2021 (CD006440.pub3) · 2021 · PubMed 33417236

Result: null (CBT vs psychoeducation); harm-direction for unsupported online CBT vs psychoeducation on mood

• 638 people
• 26 weeks
• p = .004 (HADS depression, unsupported CBT vs psychoeducation)

What was measured: psychological-distress, depression, anxiety, caregiver-strain, competence

What was found: All three arms improved from baseline on GHQ-12 (P<.001) but neither CBT arm was better than psychoeducation. CBT without telephone support was worse than psychoeducation on HADS depression (mean difference 1.86, 95% CI 0.61-3.11; P=.004) and Relative Stress Scale (MD 3.11, 95% CI 0.13-6.09; P=.04)

Study design: three-arm parallel-group online RCT; primary outcome GHQ-12 at 26 weeks; completer analysis (ANCOVA)

Strength: Moderate (large randomization, but only 208 of 638 in the analysis population)

Limitations: Only 208 of 638 randomized (33%) contributed to the primary analysis; no untreated control, so within-arm improvement cannot be attributed to the programmes

Source: Fossey et al., JAMDA 2021 · 2021 · PubMed 33288467

Result: benefit (depression, small); null (anxiety, burden, competence)

• varied
• p = 0.027 (depression)

What was measured: depression, anxiety, burden, competence

What was found: Depression Hedges g = -0.21 (95% CI -0.410 to -0.025; z=-2.216; p=0.027) favouring intervention; no significant differences for anxiety, burden or competence. Best results for multicomponent online programmes (psychoeducation + psychological skills training + professional support + online forums/support groups)

Study design: systematic review and meta-analysis (PubMed, PsycInfo, CINAHL, Web of Science, Cochrane), studies Jan 2014-Jul 2018

Strength: Moderate (10 studies; small pooled effect)

Limitations: Only 10 heterogeneous studies including quasi-experimental designs; medium-small effects despite high study quality

Source: Etxeberria, Salaberria & Gorostiaga, Aging & Mental Health 2021 · 2021 · PubMed 32363901

Result: benefit (tentative)

• varied

What was measured: depression, anxiety, burden, stress, well-being

What was found: Most studies suggested benefit to caregiver well-being including depression, anxiety and burden; strongest signal for online psychological support in several small RCTs; overall study quality low

Study design: systematic review with narrative synthesis (all study designs)

Strength: Low (mixed designs, low quality)

Limitations: Heterogeneous interventions and outcomes; no meta-analysis; search April 2018

Source: Hopwood et al., Journal of Medical Internet Research 2018 · 2018 · PubMed 29895512

Result: benefit

• 245 people
• 8 lessons with online feedback (guided self-help)
• p = 0.034 (depression); 0.007 (anxiety)

What was measured: depression, anxiety

What was found: Post-intervention symptoms lower in intervention group: depression (CES-D) P=0.034, effect size 0.26; anxiety (HADS-A) P=0.007, effect size 0.48

Study design: randomized controlled trial vs minimal intervention (e-bulletins)

Strength: Moderate (RCT; internet-recruited, well-educated sample)

Limitations: Non-completion high: 55.7% did not finish all 8 lessons (Pot et al. 2015, PMID 25648589); all data collected online

Source: Blom et al., PLoS One 2015 · 2015 · PubMed 25679228

Result: benefit (perceived stress); null (depression, bother, QoL)

• 150 people
• 3 months
• p = .017 (perceived stress)

What was measured: perceived-stress, caregiver-upset, depression, quality-of-life

What was found: Perceived stress (PSS, primary) improved significantly in the iCare arm but not the information-only arm (p=.017); no significant change in bother (RMBPC), depression (CES-D) or quality of life; roughly one-third of enrolled caregivers dropped out before completion

Study design: RCT vs education/information-only online condition

Strength: Low (single outcome significant; ~33% attrition)

Limitations: High dropout; effect size not reported in abstract; comparison arm received information so absolute effect unknown

Source: Kajiyama et al., Aging & Mental Health 2013 · 2013 · PubMed 23461355

Result: benefit (short-term, multiple outcomes); mostly not maintained at 6 months

• 81 people
• 8 weekly telephone sessions; 6-month follow-up

What was measured: depression, anxiety, physical-health, pre-death-grief, coping, emotional-well-being

What was found: At post-assessment vs control: significantly lower depressive symptoms, fewer physical symptoms, better physical health, more resource utilization, better coping and emotional well-being (the abstract does not give their size). At 6 months: less pre-death grief, fewer physical symptoms, more resource use; most other effects not maintained

Study design: RCT vs untreated control

Strength: Low (small RCT, n=81; authors call results preliminary)

Limitations: Untreated control; small sample; anxiety (co-primary) not reported as significant in abstract; effects faded by 6 months

Source: Risch, Lechner-Meichsner & Wilz, Clinical Gerontologist 2025 · 2025 · PubMed 39169671

Result: benefit (internet, telephone, individual formats); null (group, combined formats)

• 4,191 people
• varied across 37 RCTs

What was measured: depression

What was found: Vs controls: internet CBT SMD -1.33 (95% CI -2.18 to -0.66; GRADE low); telephone CBT SMD -1.29 (-1.89 to -0.61; GRADE moderate); individual CBT SMD -1.04 (-2.01 to -0.07; GRADE very low); group and combined formats not significantly effective; no statistically significant differences between the five formats

Study design: systematic review with pairwise and network meta-analysis of RCTs to 15 January 2022 (English and Chinese databases)

Strength: Low to moderate (GRADE ratings low to very low for most comparisons; 37 RCTs, 4,191 caregivers)

Limitations: Effect sizes implausibly large relative to pairwise meta-analyses (e.g., Hopkinson 2019 SMD -0.34); formats did not differ significantly from each other; authors advise cautious interpretation

Source: Sun, Ji, Leng & Wang, Journal of Affective Disorders 2022 · 2022 · PubMed 35429541

Result: benefit (depression, stress; small); null (anxiety)

• varied; subgroup analysis compared <=8 vs >8 sessions
• p = <.001 (depression, stress); .47 (anxiety)

What was measured: depression, anxiety, stress, quality-of-life

What was found: Depression SMD -0.34 (95% CI -0.47 to -0.21); stress SMD -0.36 (-0.52 to -0.20); anxiety SMD 0.10 (-0.18 to 0.39), not significant. Significant reductions limited to group (not individual) formats; <=8 sessions as effective as >8 (depression MD 0.79, 95% CI -0.45 to 2.03, P=.21 between dose groups)

Study design: systematic review and meta-analysis of CBT trials (MEDLINE, EMBASE, CINAHL, PsycINFO, Cochrane, grey literature)

Strength: Moderate (25 pooled studies; small effects; heterogeneity in formats)

Limitations: Abstract gives no pooled N; format subgroup findings (group better than individual) conflict with the 2022 network meta-analysis of CBT delivery formats (CI record on Sun et al. 2022)

Source: Hopkinson, Reavell, Lane & Mallikarjun, The Gerontologist 2019 · 2019 · PubMed 29529290

Result: benefit (short-term depression/anxiety); uncertain (burden)

• 201 people
• varied (not stated in abstract)
• p = <0.001 (depression vs active control)

What was measured: depression, anxiety, burden, coping, dropout

What was found: Vs active controls: depressive symptoms SMD -0.63 (95% CI -0.98 to -0.28; 3 trials, 135 participants); anxiety MD -7.50 (-13.11 to -1.89; 1 trial); burden SMD 0.24 (-0.11 to 0.58, possible slight increase). Vs inactive controls: depression MD -1.97 (-6.89 to 2.95, not significant)

Study design: Cochrane systematic review and meta-analysis of RCTs

Strength: Low to very low quality (GRADE)

Limitations: Only 5 small trials; high risk of detection bias; no data on QoL or institutionalization

Source: Liu et al., Cochrane Database of Systematic Reviews 2018 (CD012791.pub2) · 2018 · PubMed 30106471

Result: benefit (well-being, depression, physical symptoms, coping); well-being and coping sustained at 6 months

• 273 people
• 12 x 50-minute telephone CBT sessions within 6 months; 6-month follow-up
• p = .001 (emotional well-being, post-intervention)

What was measured: depression, emotional-well-being, physical-health, burden, coping, problem-behaviours

What was found: Post-intervention vs usual care: emotional well-being gamma=9.59 (p=.001); depression gamma=-0.23 (p=.043); physical health symptoms gamma=-0.25 (p=.019); coping with care situation gamma=0.25 (p=.005); care-recipient behaviour gamma=0.23 (p=.034). At 6-month follow-up: coping gamma=0.28 (p=.006) and 0.39 (p<.001), emotional well-being gamma=7.61 (p=.007)

Study design: randomized controlled trial vs usual care (unrestricted community resources), intention-to-treat latent change models

Strength: Moderate (RCT n=273 embedded in routine care; multiple outcomes, some marginal)

Limitations: Burden of care listed as an outcome but no significant effect reported in abstract; depression effect small and marginal (p=.043)

Source: Wilz, Reder, Meichsner & Soellner, The Gerontologist 2018 · 2018 · PubMed 29190357

Result: benefit (depression, both therapies; large at post-intervention); CBT effect persisted at follow-up, ACT did not

• 135 people
• not stated in abstract; post-intervention and follow-up assessments
• p = <.001 (depression, both arms, post-intervention)

What was measured: depression, anxiety, leisure, dysfunctional-thoughts, experiential-avoidance

What was found: Depression at post-intervention vs control: CBT d=0.98 (NNT 3.61), ACT d=1.17 (NNT 3.53), both P<.001; maintained at follow-up only for CBT (d=0.74, NNT 9.71, P=.02). Clinically significant change 26.7% CBT, 24.2% ACT, 0% control. Anxiety at post: ACT lower than CBT (d=0.50) and control (d=0.79, NNT 3.86), no follow-up effect

Study design: three-arm randomized clinical trial (CBT vs ACT vs control), caregivers with high depressive symptomatology

Strength: Moderate (single-site RCT, n=135; large post-treatment effects, attenuated at follow-up)

Limitations: Recovery at follow-up modest (10.53% CBT, 4% ACT, 0% control for depression); follow-up length not in abstract; Spanish sample

Source: Losada et al., Journal of Consulting and Clinical Psychology 2015 · 2015 · PubMed 26075381

Result: null (primary); harm (carer anxiety, stress with higher attendance)

• 488 people
• 12 weekly sessions then monthly maintenance for 7 months; 10-month primary endpoint
• p = 0.53 (primary QoL-AD); 0.04 (carer anxiety, adverse direction)

What was measured: quality-of-life, psychological-distress, caregiver-strain, cost

What was found: No differences on primary or secondary outcomes (QoL-AD mean difference 0.07, SE 0.65; F=0.48; p=0.53). Carers in the reminiscence arm reported significantly higher anxiety on a GHQ-28 subscale at 10 months (mean difference 1.25, SE 0.5; F=8.28; p=0.04); carers attending more groups showed increased caregiving stress. Economic analysis curtailed because QALY differences were negligible

Study design: multicentre pragmatic RCT vs usual care; blinded assessment at 3 and 10 months; intention-to-treat

Strength: High (large pragmatic multicentre RCT; consistent null)

Limitations: 350 of 488 dyads completed; compliance analyses are non-randomized comparisons

Source: Woods et al., Health Technology Assessment 2012 (REMCARE pragmatic multicentre RCT) · 2012 · PubMed 23211271

Result: null

• 236 people
• 15 months post-randomisation follow-up

What was measured: depression, quality-of-life

What was found: No benefit of access to a befriender facilitator on the primary outcome (HADS depression) or any secondary outcome (EuroQoL) at 15 months

Study design: single-blind RCT, community settings in East Anglia and London; intention-to-treat

Strength: Moderate (RCT n=236; clear null)

Limitations: Uptake of befriending was incomplete, so this tests offering the scheme rather than receiving befriending

Source: Charlesworth et al., BMJ 2008 (BECCA trial) · 2008 · PubMed 18505757

• Year: 2024

Who this describes: Medicare fee-for-service beneficiaries with dementia and their caregivers

Official program description. Established Program Track delivered services from July 1, 2024; New Program Track had a pre-implementation year and began services July 1, 2025.

Source: CMS Innovation Center GUIDE Model page and Model Overview Fact Sheet · 2024

• Year: 2024

Who this describes: Qualifying caregivers of GUIDE-aligned beneficiaries

Official program parameter. Respite is 'not currently a Medicare-covered service outside of the hospice benefit' (fact sheet).

Source: CMS Innovation Center GUIDE Model page · 2024

• Year: 2024

Who this describes: GUIDE participants (Medicare Part B enrolled providers/suppliers)

Official. Five patient tiers: with caregiver - low complexity (mild dementia), moderate complexity (moderate/severe dementia, low-moderate caregiver strain), high complexity (moderate/severe dementia, high caregiver strain); without caregiver - low complexity (mild) and moderate-to-high complexity (moderate/severe). Dollar amounts of PBPM payments not on loaded pages.

Source: CMS GUIDE Model Overview Fact Sheet (PDF) · 2024

• Year: 2024

Who this describes: GUIDE participants

Official. Eligible beneficiaries: traditional Medicare Parts A and B, dementia diagnosis, not in Medicare Advantage, PACE, hospice, or long-term nursing home (LeadingAge summary).

Source: CMS Innovation Center GUIDE Model page · 2024

• Year: 2026

Who this describes: GUIDE-aligned beneficiaries and caregivers

Stated aims (delay nursing home placement, reduce hospital/ED use, reduce caregiver burden, improve QoL, reduce Medicare/Medicaid spending) are goals, not results. Related RCT evidence: Care Ecosystem (benefit) and D-CARE (not given) in interventions file.

Source: CMS Innovation Center GUIDE Model page · 2026

• Year: 2015-2016 (published 2018)

Who this describes: Adults providing unpaid care to a person with dementia in 38 states, the District of Columbia, and Puerto Rico

PMID 30138069; PMC6137771. (cost-replacement method applied to 2015/2016 BRFSS caregiver-module respondents, valued using US Bureau of Labor Statistics May 2016 occupation wage profiles). Covers 38 states + DC + Puerto Rico, not all 50 states, and a different, earlier vintage than the site's national $446.3 billion modelled Facts and Figures figure above, which uses a different (survey-based, scaled to Census counts) method -- not directly comparable to it.

Source: Rabarison KM, Bouldin ED, Bish CL, et al. The Economic Value of Informal Caregiving for Persons With Dementia: Results From 38 States, the District of Columbia, and Puerto Rico, 2015 and 2016 BRFSS. Am J Public Health 2018;108(10):1370-1377 · 2015-2016 (published 2018)