Complex interventions for improving independent living and quality of life amongst community-dwelling older adults: a systematic review and meta-analysis

Abstract

Background

community-based complex interventions for older adults have a variety of names, including Comprehensive Geriatric Assessment, but often share core components such as holistic needs assessment and care planning.

Objective

to summarise evidence for the components and effectiveness of community-based complex interventions for improving older adults’ independent living and quality of life (QoL).

Methods

we searched nine databases and trial registries to February 2022 for randomised controlled trials comparing complex interventions to usual care. Primary outcomes included living at home and QoL. Secondary outcomes included mortality, hospitalisation, institutionalisation, cognitive function and functional status. We pooled data using risk ratios (RRs) or standardised mean differences (SMDs) with 95% confidence intervals (CIs).

Results

we included 50 trials of mostly moderate quality. Most reported using holistic assessment (94%) and care planning (90%). Twenty-seven (54%) involved multidisciplinary care, with 29.6% delivered mainly by primary care teams without geriatricians. Nurses were the most frequent care coordinators. Complex interventions increased the likelihood of living at home (RR 1.05; 95% CI 1.00–1.10; moderate-quality evidence) but did not affect QoL. Supported by high-quality evidence, they reduced mortality (RR 0.86; 95% CI 0.77–0.96), enhanced cognitive function (SMD 0.12; 95% CI 0.02–0.22) and improved instrumental activities of daily living (ADLs) (SMD 0.11; 95% CI 0.01–0.21) and combined basic/instrumental ADLs (SMD 0.08; 95% CI 0.03–0.13).

Conclusions

complex interventions involving holistic assessment and care planning increased the chance of living at home, reduced mortality and improved cognitive function and some ADLs.

aged, Geriatric Assessment, independent living, quality of life, Community Health Services, systematic review, older people

Key Points

• Community-based complex interventions for older adults have heterogeneous components.

• Most community-based complex interventions for older adults involved holistic assessment and care planning.

• Nurses were the most frequent care coordinators in multidisciplinary care.

• Complex interventions increased the chance of living at home, but not quality of life, amongst community-dwelling older adults.

• Complex interventions also reduced their mortality, enhanced cognitive function and improved some activities of daily living.

Background

The world’s population is ageing rapidly [1]. Although the speed and pattern of population ageing vary by country, the growing proportion of older adults challenges hospital-centric healthcare systems [2]. Hospital admission is expensive, and the focus of most hospital care on single conditions is poorly aligned with the needs of older adults with multimorbidity, polypharmacy and frailty [3]. In hospital settings, a range of complex interventions has been developed to meet the care needs of older adults, including Comprehensive Geriatric Assessment (CGA), other kinds of discharge planning and more complex reorganisations of care [4]. CGA takes a multidisciplinary approach to a holistic assessment of needs, with coordinated health and social care to address those needs. Although there is evidence that CGA is an effective intervention in hospital inpatients [5], the evidence of effectiveness in the community is less clear.

Community-based complex interventions decrease the risk of unplanned hospital admissions amongst older adults at risk of poor health outcomes [6], and there is some evidence they improve quality of life (QoL) and reduce caregiver burden [7]. However, previous reviews have not evaluated other critical outcomes regarding independent living, such as living at home and institutionalisation [6, 7]. Additionally, although reviews often focus on how researchers classify or name their interventions (e.g. in reviews of ‘CGA’), interventions with the same name are frequently heterogeneous in their intervention components, whereas interventions with different names often share core components [8]. Such heterogeneity may influence the adoption of evidence and hence the formulation of health and social care policies.

This systematic review with meta-analysis, therefore, aims to summarise current evidence on the effectiveness of community-based complex interventions (irrespective of how they are named) intended to improve independent living and QoL of older adults.

Methods

Full methods are reported in Box 1, Supplementary file, and briefly reported here. The review protocol was registered in PROSPERO (CRD42021274017). Eligible studies were randomised controlled trials (RCTs) conducted in high-income countries which recruited community-dwelling adults that either explicitly targeted older adults or where the mean participant age was ≥65 years. Community-dwelling was defined as living independently at home (including in extra-care housing but excluding care/nursing home residents) regardless of the need for care assistance.

Interventions

Complex interventions include several interacting components [9], which we classified in terms of the Taxonomy of Health Systems Interventions published by the Cochrane Effective Practice and Organisation of Care [10] and the NICE (National Institute for Health and Care Excellence) multimorbidity guideline document [8] (Table S1). RCTs where the only intervention was health education workshops or group activities without individual assessment or delivery of care to individuals were excluded.

Comparators

The comparator was ‘usual care’ in the setting the study was based in. RCTs offering minor enhancements to usual care in the control arm, such as written educational materials, were also eligible if they explicitly stated the content of additional components.

Outcomes

The primary outcomes examined were living at home and QoL. Living at home was defined either as a reported outcome or the inverse of mortality and institutionalisation (admission to a care or nursing home) combined at the end of follow-up. QoL had to be measured by validated self-reported outcome instruments (any of Short Form (SF)-12, SF-36, EQ-5D-3L, EQ-5D-5L, 15D, QUAL-E and Cantril’s Ladder). Secondary outcomes included mortality, hospitalisation (≥1 during follow-up), institutionalisation (≥1 during follow-up), cognitive function (measured by validated instruments) and functional status (measured by validated assessments of activities of daily living (ADLs), instrumental activities of daily living (IADLs), combined ADLs/IADLs or physical mobility).

Search strategy and selection criteria

Six electronic databases (MEDLINE, Embase, CINAHL, PsycINFO, Web of Science, Cochrane Library) and three trial registries (ClinicalTrials.gov, ISRCTN, ICTRP) were searched from inception to February 2022. Search strategies are defined in Box 2, Supplementary file, with additional hand-searching of reference lists. Covidence (https://www.covidence.org/) was used for data management, with title, abstract and full-text screening done by two independent reviewers. Discrepancies were resolved by discussion and, if necessary, involvement of a third reviewer.

Data extraction, risk of bias assessment and quality of evidence assessment

Characteristics and outcome data of the eligible studies were extracted by a single reviewer with validation by a second reviewer, using the pre-specified data extraction sheet. Risk of bias assessment was conducted for all included studies using the Cochrane risk-of-bias tool for RCTs-2 [11]. The Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach was adopted for the assessment of the overall quality of evidence of meta-analyses [12].

Data synthesis

Meta-analysis

Dichotomous outcomes, including living at home, mortality, hospitalisation and institutionalisation were synthesised using risk ratios (RRs) and continuous outcomes, including QoL, cognitive function and functional status were pooled using standardised mean differences (SMDs). A fixed-effects model was used when the heterogeneity was low (I² < 30%), and a random-effects model otherwise [13]. For RCTs with multiple periods of follow-up, only the outcome results from the longest follow-up were pooled in meta-analyses [13], and only results from intention-to-treat analyses were synthesised [14]. Results from per-protocol analyses and other unpooled results are shown in Table S2, with unpooled results also synthesised narratively.

Small-study effects were examined using funnel plots and Egger’s regression tests [15]. For the primary outcomes, sensitivity analyses included leave-one-out analysis to explore whether findings were driven by single studies [16] and comparison of pooled results between studies at low/moderate versus high risk of bias [13]. For continuous outcomes, further sensitivity analysis compared pooled results between studies reported in change score from baseline against those reported in follow-up score [17]. For meta-analyses with ≥10 studies [13, 18], subgroup analyses were conducted stratified by: length of follow-up (short- versus medium- versus long-term follow-up); location of intervention delivered (home-only versus non-home settings ± home); frailty, disability or functional decline of participants (present versus absent); multidisciplinary care (scheduled versus not); home/telephone follow-up (scheduled versus not); and self-management (planned versus not).

Results

Study selection and characteristics

In total, 18,714 unique records were screened, 333 full texts assessed and 50 RCTs conducted between 1984 and 2019 were included in this review (Figure 1). The list of included articles is shown in Box 3, Supplementary file. Fifteen (30%) RCTs took place in the European Union, 14 (28%) in the United States and 3 (6%) in the United Kingdom (Table 1). The majority (n = 37; 74%) of studies adopted frailty, disability or functional decline as an inclusion criterion. Twenty-nine (58%) studies involved interventions provided in day hospitals, general practice surgeries or other health and community care providers. The duration of intervention ranged from 10 weeks to 48 months. A total of 31,659 participants were involved in the studies, with the average age (mean or median) ranging from 69.5 to 86.3 years.

Figure 1

Flow of literature search and selection. ^*One RCT was reported across two included papers [19, 20]

Table 1

Characteristic of included randomised controlled trials

First author (publication year)	OECD country	Location of intervention delivery	Inclusion criteria	Exclusion criteria	Duration of intervention	Sample size	Participant age mean (SD)
Beland (2006)	Canada	Home Health care providers Community care providers	Older than 64 years old; Community-dwelling; Residing within the two local community services centre territories; Being competent in French or English (participant or caregiver); Having a participating caregiver (if a caregiver existed); and Having at least moderate disability, defined by the Functional Autonomy Measurement System scale	Having pending nursing home admission; or Moving out of the local community services centre territories	22 months	I: 606 C: 624	I: Median 82.0; Range 74.0–104.0 C: Median 82.0; Range 64.0–104.0
Bernabei (1998)	Italy	Home Health care providers Community care providers	Aged 65 and over; and Receiving home health services or home assistance, usually because of multiple geriatric conditions	Not reported	12 months	I: 99 C: 100	I: 80.7 (7.1) C: 81.3 (7.4)
Bleijenberg (2016)	Netherlands	Home	Potentially frail individuals aged 60 and older	Terminal illness or estimated life expectancy of ≤3 months; or Living in assisted-living facilities or nursing homes	12 months	I: 790 (Screening group) I: 1,446 (Screening + Nurse-Led Care group) C: 856	I: 73.5 (8.2) (Screening group) I: 74.0 (8.2) (Screening + Nurse-Led Care group) C: 74.6 (8.8)
Blom (2016)	Netherlands	GP surgeries	Aged ≥75; Community-dwelling; and Complex problems (3+ from: (i) Poor outcomes on disability; (ii) Feelings of loneliness; (iii) Poor health-related QoL; or (iv) Less GP contact time)	Terminal illness or estimated life expectancy of ≤3 months	12 months	I: 288 C: 1,091	I: Median 82.0; IQR 78.8–86.9 C: Median 83.7; IQR 79.8–88.0
Boult (2011)	United States	Home Health care providers Community care providers	65 years or older; and High risk of generating high healthcare expenditures during the following year, estimated by the Hierarchical Condition Category predictive model	Not reported	20 months	I: 404 C: 446	I: 77.8; Range 66.0–96.0 C: 77.1; Range 66.0–106.0
Boult (2013)	United States	Home Health care providers Community care providers	65 years or older; and High risk of generating high healthcare expenditures during the following year, estimated by the Hierarchical Condition Category predictive model	Did not have a telephone; Did not speak English; Planning extended travel; Failed the brief cognitive screen; or Did not have a proxy who could provide consent	32 months	I: 485 C: 419	I: 77.2 (SD not reported) C: 78.1 (SD not reported)
Brettschneider (2015)	Germany	Home	Older than 80 years; Residing in Leipzig or Halle; and Lived at home or planned discharge to home	Had insufficient German language skills; Suffered from cognitive impairment; Unable to give informed consent; or Had a care level > 1, according to German long term care insurance	18 months	I: 133 C: 145	I: 84.9; 3.5 C: 84.7; 3.5
Burns (2000)a	United States	Multidisciplinary clinics for the study	65 years of age or older; Admitted to either the medical, surgical, or neurology services at the Memphis Veterans Affairs Medical Centre; and Having two or more of the following: (i) ≥1 ADL deficits; (ii) ≥2 chronic medical conditions); (iii) ≥2 acute care hospitalisations in the previous year; or (iv) ≥6 scheduled prescription drugs	Having admitted to nursing home; Required inpatient Geriatric Evaluation and Management and/or rehabilitation during hospital stay; Having a terminal illness with a life expectancy of <6 months; Having moderate to severe dementia (MMSE score < 18); Having end stage disease; or Chose not to participate in the study	24 months	I: 60 (Analysed: 49) C: 68 (Analysed: 49)	I: 71.7; 6.3 C: 70.8; 3.7
Buss (2016)a	Germany	Home	Aged >60 years; Ability to communicate motorically, cognitively, and psychologically; Ability to communicate in German; Residing in Hamburg; and Having functional mobility impairment of the musculoskeletal system or stroke	Deficient orientation to place or time; Cognitive impairment (MMSE score < 25); or Receiving palliative care	12 months	I: 32 (Analysed: 24–27) C: 33 (Analysed: 28–31)	I: 81.8; 8.4 C: 83.0; 7.5
Caplan (2004)b	Australia	Home Health care providers Community care providers	Aged 75 or older; and Discharged home from the emergency department	Living in nursing home; Had previously already been enrolled in this study; or Living out of the local area of the hospital	18 months	I: 370 (Analysed: 293–370) C: 369 (Analysed: 282–369)	I: 82.1; 6.6 C: 82.4; 5.2
Coleman (1999)	United States	Home	65 years of age or older; and At high risk of hospitalisation and functional decline, defined by validated computer-based predictive index	Too ill to participate; Having moderate to severe dementia; Residence in a nursing home; Having terminal illness; or Disenrolled	24 months	I: 96 C: 73	I: 77.3 (SD not reported) C: 77.4 (SD not reported)
Counsell (2007)	United States	Home	Age 65 years or older; At least 1 visit to a primary care clinician at the same site within the past 12 months; and Having income less than 200% of the federal poverty level	Residence in a nursing home; Living with a study participant already enrolled in the trial; Enrolled in another research study; Receiving dialysis; Severe hearing loss; English-language barrier; No access to a telephone; Severe cognitive impairment, defined by SPMSQ score; or Without an available caregiver to consent to participate	24 months	I: 474 C: 477	I: 71.8; 5.6 C: 71.6; 5.8
Di Pollina (2017)	Switzerland	Home Day hospitals	60 years and older; and Presence of frailty, defined by the Resident Assessment Instrument-Home Care	Did not meet frailty criteria; or Could not speak French	36 months	I: 122 C: 179	I: 81.8; 8.2 C: 81.9; 8.2
Dolovich (2019)	Canada	Home Health care providers Community care providers	Aged 70 years or older; and Living in Hamilton	Away for more than 50% of trial duration; Receiving long-term/palliative care; or Neither they nor their family member spoke English	6 months	I: 158 C: 154	I: 78.1; 6.3 C: 79.06; 6.6
Engelhardt (1996)	United States	Day hospitals	Aged 55 or older; Were above average users (10 or more clinic visits) of Department of Veterans Affairs Medical Centre outpatient clinic services in the previous 12 months; and With at least two impairments on ADLs or IADLs	Hospitalised with a psychiatric diagnosis within the previous year; 2. With severe cognitive impairments as assessed by SPMSQ; 3. Received care within a year before screening from: (i) Oncology or renal clinics; (ii) Hospital-based home care; (iii) Adult health day care; or (iv) an Inpatient Geriatric Evaluation and Management Unit; Residing in nursing homes; or Reported receiving the majority of care in the previous year from non-veterans affairs providers	16 months	I: 80 C: 80	I: 71.7; 6.8 C: 72.6; 5.8
Fabacher (1994)a	United States	Home	Veterans of the United States armed services; Age 70 years or older; Not currently enrolled in a Veterans Affairs outpatient clinic; and Not suffering from a known terminal disease or dementia	Not reported	12 months	I: 131 (Analysed: 100) C: 123 (Analysed: 95)	I: 73.5; 4.3 C: 71.8; 7.0
Ford (2019)a	United Kingdom	GP surgeries	Aged 18 or over; and Identified as in the top 2% for risk of unplanned admission and diagnosed with at least two of 40 morbidities in Barnet’s analysis of multimorbidity	Deemed to be unable to participate in goal-setting in the G’s professional opinion; Had received a care planning consultation in the previous 3 months; or Required translation services to communicate verbally	6 months	I: 24 (Analysed: 18) C: 28 (Analysed: 23)	I: 80.4; 8.7 C: 77.2; 9.4
Fristedt (2019)	Sweden	Home Health care providers	Aged ≥75 years; Community-dwelling; Having more than 3 chronic diagnoses; Prescribed 6 or more pharmaceutical drugs for continuous use; and With >3 hospital stays (>24 hours in hospital) during the last 6 month	Deceased; Lived in a nursing home; and Had a hospital admission not relevant to the Mobile Geriatric Team concept	12 months	I: 31 C: 31	I: 84.0; 5.1 C: 86.0; 5.7
Gitlin (2006)	United States	Home	Aged 70 and older; Cognitively intact (MMSE >23); English speaking; Not receiving home occupational therapy or physical therapy; and Functionally vulnerable (needing help with two IADLs, having difficulty performing one ADL, or experiencing one or more falls within 1 year before study entry)	Totally dependent; Homebound; or Receiving services to address functional problems	12 months	I: 160 C: 159	I: 79.5; 6.1 C: 78.5; 5.7
Godwin (2016)a	Canada	Home	80 years or older; and Functioning well cognitively and living independently in the community	Living in a nursing home; Not able to give informed consent; MMSE score < 25; Had profound communication difficulties; or Receiving in-home care without which they would require admission to a nursing home	12 months	I: 121 (Analysed: 95) C: 115 (Analysed: 86)	I: 85.3; 4.5 C: 85.7; 3.6
Hendriksen (1984)	Denmark	Home Health care providers Community care providers	Aged 75 years or more; and Living in a suburb (Roedovre municipality) of Copenhagen	Not reported	36 months	I: 285 C: 287	I: Median 78.4; Range 75–96 C: Median 78.6; Range 75–95
Hoogendijk (2016)	Netherlands	Home Health care providers Community care providers	Aged 65 or over; Community-dwelling; and Had moderate to severe disability, as defined by PRISMA-7 score of 3 or more	Not reported	24 months	I: 1,147 C: 1,147	I: 80.5; 7.5 C: 80.5; 7.5
Kerse (2014)a	New Zealand	GP surgeries	Aged 75 years and older; Community-dwelling; and Participating primary care practices	Living in residential care; Receiving palliative care; or Terminally ill	36 months	I: 2,049 (Analysed: 1,553–2,049) C: 1,844 (Analysed: 1,428–1,844)	I: 80.4; 4.6 C: 80.3; 4.5
Kono (2012)	Japan	Home Health care providers Community care providers	Aged 65 years or older; Certified as Support Level 1 or 2 in the Long-Term Care Insurance; Living at home at the baseline survey; and Not having utilised formal long-term care services, which are reimbursed by the Long-Term Care Insurance, for the past 3 months	Not reported	24 months	I: 161 C: 162	I: 80.3; 6.7 C: 79.6; 6.4
Lewin (2013)	Australia	Home	≥65 years of age; Referred for personal care; Not having a diagnosis of dementia or other progressive neurological disorders; Not receiving palliative care; and Able to communicate in English	Not reported	12 months	I: 375 C: 375	I: 82.7; 7.7 C: 81.8; 7.2
Lihavainen (2012)	Finland	Home Health care providers Community care providers	75-year-old and older; and Residing in Kuopio, Eastern Finland	Refused to participate; or Moved out of the area	24 months	I: 404 C: 377	I: 81.0; 4.9 C: 81.1; 5.1
Liimatta (2019)	Finland	Home	75 years old or older; Home dwelling; Not receiving home help or nursing services; Finnish speaking; and Living permanently in the Hyvinkää area	Not reported	24 months	I: 211 C: 211	I: 80.8; 4.3 C: 81.8; 4.3
Markle-Reid (2010)a	Canada	Home Health care providers Community care providers	Aged 75 years and older; Newly referred to and eligible for home support services through the Community Care Access Centres; Living in the community (not in a nursing home or other long-term care facility); Mentally competent to give informed consent; and Competent in English or with a translator available	Not reported	6 months	I: 54 (Analysed: 49) C: 55 (Analysed: 43)	I: 75–85 (n = 28) I: 86 or older (n = 21) C: 75–85 (n = 22) C: 86 or older (n = 21)
Markle-Reid (2006)a	Canada	Home Health care providers Community care providers	75 years of age or older; and Newly referred to and eligible for personal support services through the Community Care Access Centres	Refused to give informed consent; Unable to understand English; or Deemed eligible for nursing services	6 months	I: 144 (Analysed: 120) C: 144 (Analysed: 122)	I: 75–85 (n = 90) I: 86 or older (n = 30) C: 75–85 (n = 78) C: 86 or older (n = 44)
Metzelthin (2015)	Netherlands	Home Health care providers Community care providers	Aged 70 years or older; Community-dwelling; and Frail, as defined by a Groningen Frailty Indicator of 5 or higher	Terminally ill; Confined to bed; Had severe cognitive or psychological impairments; or Unable to communicate in Dutch	24 months	I: 193 C: 153	I: 77.5; 5.3 C: 76.8; 4.9
Newbury (2001)a	Australia	GP surgeries	Aged 75 years or over; and Living independently in the community	Not reported	12 months	I: 50 (Analysed: 45) C: 50 (Analysed: 44)	I: Median 80; Range 75–91 C: Median 78.5; Range 75–88
Parsons (2017)	New Zealand	Home	Aged ≥65 years; and At high risk of permanent institutional care by the regional assessment agency (dementia, with associated behavioural problems; incontinence; carer stress; repeated falls; or frailty)	Needed immediate placement in residential care; or Inability to communicate in English	24 months	I: 56 C: 57	I: 82.7; 7.3 C: 83.5; 7.6
Parsons (2013)	New Zealand	Home	Older than 65 years (55 years if Maori or Pacific Islander); Community-dwelling; and New referral for home care	Severe cognitive impairment (Abbreviated Mental Test score of less than 7); or Referral for assessment for admission to a residential facility, carer support, or short-term services	6 months	I: 108 C: 97	I: 79.1; 6.9 C: 76.9; 7.6
Ploeg (2010)b	Canada	Home Health care providers Community care providers	Aged 75 years or older; They or their proxy is able to answer questions in English; and Resided in the city of Hamilton	Received home care services; Lived in a nursing home or long-term care home; Identified by family physician as needing palliative care; Scheduled for major elective surgery in the next year; or Planning to leave the country for more than one month during the 12-month follow-up period	12 months	I: 361 (Analysed: 331–361) C: 358 (Analysed: 314–358)	I: 81.0; 4.1 C: 81.3; 4.4
Radwany (2014)	United States	Home	≥60 years of age; Newly enrolled in PASSPORT (Ohio’s community-based, long-term care Medicaid waiver programme); Eligible for both Medicare and Medicaid benefits; Passed mental status screening (Mental Status Questionnaire); and Had one of the following: (i) Congestive heart failure and being actively treated; (ii) Chronic obstructive pulmonary disease and on home oxygen; (iii) Diabetes with renal disease, neuropathy, visual problems, or coronary artery disease; (iv) End-stage liver disease or cirrhosis; (v) Cancer (active, not history of) except skin cancer; (vi) Renal disease and actively receiving dialysis; (vii) Amyotrophic lateral sclerosis with history of aspiration; (viii) Parkinson’s disease stages 3 and 4; or (ix) Pulmonary hypertension	Active alcoholics (drink ≥2 drinks per day on average); Illegal substance users; Had schizophrenia or psychotic; Unable to pass the Mental Status Questionnaire; or Already enrolled in hospice	12 months	I: 40 C: 40	I: 69.5 (SD not reported) C: 68.8 (SD not reported)
Reuben (1999)a	United States	GP surgeries	65 years of age or older; Community-dwelling; and Had failed a screen for at least one of four conditions (falls, urinary incontinence, depressive symptoms or functional impairment)	Did not speak English; Did not have a telephone; Did not have a primary care physician; Were demented or had MMSE <24; or Had other mental, emotional, or physical disorders to the extent that they could not be expected to complete the questionnaires and protocol required for the study	15 months	I: 180 (Analysed: 176) C: 183 (Analysed: 175)	I: 75.8; 6.1 C: 75.9; 5.7
Rosstad (2017)	Norway	Home	70 years or older; and Served by one of the included clusters or scheduled to receive home care services (because of functional and/or cognitive impairment) after discharge from hospital	Caregivers or health personnel responsible for the care services (cognitive impairments)	12 months	I: 163 C: 141	I: 83.1; 5.7 C: 82.4; 5.7
Sahlen (2006)	Sweden	Home	75 years and older; Healthy	Not reported	24 months	I: 248 C: 346	I: 79.7; 3.9 C: 79.8; 4.3
Salisbury (2018)	United Kingdom	GP surgeries	Aged 18 years or older; and With at least three types of chronic condition	Had a life expectancy of less than 12 months; At serious suicidal risk; Known to be leaving the practice within 12 months; Unable to complete questionnaires in English; Taking part in another healthcare research project; Lacked the capacity to give consent (in Scotland only, for legal reasons); or Unsuitable to be invited for other reasons, deemed by GPs	15 months	I: 797 C: 749	I: 71.0; 11.6 C: 70.7; 11.4
Shapiro (2002)	United States	Home Health care providers Community care providers	Aged 60 or older; On a waiting list to receive social services through the State of Florida’s Community Care for the Elderly programme; and Characterised as ‘moderate risk’ based on a uniform state-wide assessment device (based on chronic health conditions, ADL limitations and other measures of physical and psychological impairment)	Moved out of the moderate-risk classification; Died; Unable to be contacted by telephone; Unable to self-report; or Institutionalised	18 months	I: 40 C: 65	I: 77.7 (SD not reported) C: 77.1 (SD not reported)
Sherman (2016)	Sweden	Home	75 years old; Living at home; and Registered at the participating healthcare centres	Not reported	12 months	I: 176 C: 262	I: 75 (SD not reported) C: 75 (SD not reported)
Spoorenberg (2018)	Netherlands	Home Health care providers Community care providers	Aged 75 and over; Living at home or in a home for the elderly; and INTERME-D-E-SA ≥16 (Complex care needs)/INTERMED-E-SA <16 and Groningen Frailty Indicator ≥5 (Frail)/INTERMED-E-SA <16 and Groningen Frailty Indicator <5 (Robust)	Long-term admission to a nursing home (not just for rehabilitation); Receiving an alternative type of integrated care; or Participating in another research study	12 months	I: 747 C: 709	I: 80.6; 4.5 C: 80.8; 4.7
Stuck (1995)b	United States	Home	75 years of age or older; and Living at home	Had severe cognitive impairment; Had language problems; Planned to move to a nursing home; Planned to move away; Self-reported terminal disease; Participated in another randomised trial; or Had severe functional impairment	36 months	I: 215 C: 199	I: 81.0; 3.9 C: 81.4; 4.2
Suijker (2016)	Netherlands	Home Health care providers Community care providers	Aged 70 years and over; and At risk of functional decline (ISAR-PC score ≥ 2)	Had a life expectancy of less than three months; Suffered from dementia; Did not understand Dutch; Planned to move or spend a long-time abroad; or Lived in a nursing home	12 months	I: 1,209 C: 1,074	I: Median 82.6; Range 76.8–86.8 C: Median 82.9; Range 77.3–87.3
Szanton (2011)a	United States	Home	Aged 65 and older; Cognitive function MMSE score ≥ 24; Reported difficulty with one or more ADLs or two or more IADLs; Low income (house-hold income 199% of the Federal Poverty Level); and Be able to stand with or without assistance	Hospitalised more than 3 times in the previous year; Receiving in-home rehabilitation (nursing, physical therapy, or occupational therapy); Had a terminal diagnosis with less than 1 year expected survival as determined by their physician; Receiving active cancer treatment; Had plans to move in less than 1 year; or Not competent to provide informed consent	6 months	I: 24 C: 16	I: 79.0; 8.2 C: 77.0; 7.1
Thomas (2007)	Canada	Home Health care providers Community care providers	Aged 75 years or older; Living at home; Not receiving formal home care; and Able to identify an informal caregiver	Not reported	48 months	I: 175 (Intervention 1 group); 170 (Intervention 2 group) C: 175	I: 80.7; 4.3 (Intervention 1 group) I: 80.4; 4.4 (Intervention 2 group) C: 80.7; 4.5
Tuntland (2015)a	Norway	Home	Aged 18 years or older; Home-dwelling; Living in the municipality of Voss in Western Norway; Able to understand Norwegian; and Had a functional decline in one or more ADLs	In need of institution-based rehabilitation or nursing home placement; Terminally ill; or Moderately or severely cognitively reduced (subjectively assessed by health-care providers based on observation and communication	10 weeks	I: 31 (Analysed: 25–28) C: 30	I: 79.9; 10.4 C: 78.1; 9.8
van Hout (2010)	Netherlands	Home Health care providers Community care providers	Aged 75 years and older; Listed as primary care practice patient; Living at home; and Frail (self-reported score in the worst quartile of at least 2 of 6 COOP/WONCA charts: overall health ≥4; physical fitness ≥5; changes in health ≥4; daily activities ≥4; mental health ≥3; social activities ≥3)	Terminally ill as determined by primary care physicians; With dementia symptoms (self-report of memory deterioration and MMSE <24 or 7-minute screen >50%); Living in residential homes; or Participating in other research projects	18 months	I: 331 C: 320	I: 81.3; 3.9 C: 81.5; 4.3
Walters (2017)	United Kingdom	Home	Aged ≥65 years; Registered with a participating general practice; Scoring as ‘mildly frail’ on the Rockwood Clinical Frailty Scale; Community-dwelling (including extra care housing); A life expectancy of >6 months; Capacity to consent to participate (including those with dementia or communication difficulties who retained capacity)	Living in care homes; Had moderate to severe frailty or who are not frail (according to the Rockwood Clinical Frailty Scale); Were on the GP register for palliative care or dementia; Were housebound; Were already case managed; Were lacking capacity to consent; Inappropriate to have an invitation to participate for at this time, as judged by their GP	6 months	I: 26 C: 25	I: 80.38; 6.89 C: 79.68; 6.36
Zimmer (1985)a	United States	Home	Home-bound; Wishing to remain at home; Having significant illness (not primarily psychiatric) requiring medical care; Not having a physician who would make home visits; Living within Monroe County; Having a family member or friend (‘caretaker’) who could assist in their care at home; and Willing to participate	Not reported	6 months	I: 82 C: 76	I: 73.8 (SD not reported) C: 77.4 (SD not reported)

C, control group; I, intervention group; INTERMED-E-SA, The INTERMED for the Elderly Self-Assessment; ISAR-PC, Identification of Seniors At Risk—Primary Care; MMSE, Mini-Mental State Examination; OECD, Organisation for Economic Co-operation and Development; SPMSQ, Short Portable Mental Status Questionnaire.

^aStudies adopted per-protocol analysis for all outcomes.

^bStudies adopted per-protocol analysis for some outcomes (further detailed in Table S2).

Descriptions of interventions and comparators

Intervention components for each included study are summarised in Table 2, with details documented in Table S3. Forty-seven (94%) RCTs reported using holistic assessment (non-disease-focused) as one of their intervention components, and 45 (90%) studies included care planning, including multidisciplinary care plans, self-management plans or developing care plans for routine primary care management. Amongst the 27 RCTs reporting multidisciplinary care, 8 (29.6%) had their interventions delivered mainly by primary care teams without the involvement of geriatricians or other specialist clinicians, 12 (44.4%) mainly by secondary care teams without the involvement of primary care professionals and 7 (25.9%) by primary and secondary care teams (Table 3). Over half of the studies involved nurses or advanced practice nurses (APNs; n = 20; 74.1%), general practitioners (GP; n = 15; 55.6%) and/or physiotherapists (n = 14; 51.9%) as care coordinators. In 18 (66.7%) studies, nurses or APNs were responsible for coordinating the multidisciplinary care. Fourteen (28%) and eight (16%) studies provided their participants with home and telephone follow-up only, respectively, with 10 (20%) others providing both. Finally, a total of 16 RCTs reported the adoption of planned self-management as an intervention component (Table 2).

Table 2

Intervention components of the included randomised controlled trials

Study	Component of intervention
	Holistic assessment	Multidisciplinary care	Care plan development	Home follow-up	Telephone follow-up	Self-management
Beland (2006)	●	●	●
Bernabei (1998)	●	●	●	●
Bleijenberg (2016)	●a	●	●a	●a
Blom (2016)	●		●
Boult (2011)	●		●		●	●
Boult (2013)	●		●		●	●
Brettschneider (2015)	●		●	●		●
Burns (2000)	●	●	●		●
Buss (2016)			●	●	●	●
Caplan (2004)	●	●	●
Coleman (1999)	●		●			●
Counsell (2007)	●	●	●	●	●
Di Pollina (2017)	●	●	●
Dolovich (2019)	●	●	●	●
Engelhardt (1996)	●	●	●
Fabacher (1994)	●		●	●
Ford (2019)		●	●
Fristedt (2019)	●	●	●	●
Gitlin (2006)				●	●
Godwin (2016)	●		●	●
Hendriksen (1984)	●
Hoogendijk (2016)	●	●	●	●	●
Kerse (2014)	●	●	●
Kono (2012)	●	●	●	●	●
Lewin (2013)	●	●	●		●	●
Lihavainen (2012)	●	●	●
Liimatta (2019)	●		●			●
Markle-Reid (2006)	●	●	●			●
Markle-Reid (2010)	●	●	●	●	●	●
Metzelthin (2015)	●	●	●	●	●
Newbury (2001)	●		●
Parsons (2013)	●	●	●
Parsons (2017)	●		●	●
Ploeg (2010)	●	●	●	●	●
Radwany (2014)	●		●		●
Reuben (1999)	●		●		●	●
Rosstad (2017)	●		●
Sahlen (2006)	●				●
Salisbury (2018)	●		●
Shapiro (2002)	●	●	●		●
Sherman (2016)	●			●
Spoorenberg (2018)	●b	●b	●b	●b		●
Stuck (1995)	●		●	●		●
Suijker (2016)	●	●	●	●		●
Szanton (2011)	●	●	●			●
Thomas (2007)	●
Tuntland (2015)	●	●	●	●		●
van Hout (2010)	●		●	●	●
Walters (2017)	●		●	●	●	●
Zimmer (1985)	●	●	●	●
Number (%) of studies with each component	47 (94)	27 (54)	45 (90)	24 (48)	18 (36)	16 (32)

^aOnly applicable to Screening + Nurse-Led Care group.

^bOnly applicable to Complex Care Needs group and Frail group.

Table 3

Composition of multidisciplinary teams in the included randomised controlled trials

Study	Level of care team	Involved health care or social care professionals													Case coordinator(s)
		Nurse	APN	GP	Geriatrician	Other specialist doctors	Dentist	Social worker	Physiotherapist	Occupational therapist	Dietitian or nutritionist	Speech therapist	Pharmacist	Psychologist
Beland (2006)	Secondary	●		●				●	●	●			●		N or SW
Bernabei (1998)	Primary and Secondary	●		●	●			●							Details not reported
Bleijenberg (2016)a	Primary		●	●											APN
Burns (2000)	Primary		●	●				●					●	●	APN or GP or SW or Psy
Caplan (2004)	Secondary	●	●	●	●	●		●	●	●					N
Counsell (2007)	Primary and Secondary		●	●				●							APN and SW
Di Pollina (2017)	Secondary			●	●			●	●	●	●			●	N
Engelhardt (1996)	Secondary		●		●			●							G
Fristedt (2019)	Secondary	●			●				●	●					G
Hoogendijk (2016)	Primary and Secondary		●	●	●				●				●		APN and G
Kerse (2014)	Primary and Secondary		●	●	●			●	●	●					APN and GP
Kono (2012)	Primary and Secondary	●		●				●							APN and SW and CM
Lewin (2013)	Secondary	●							●	●					N and PT and OT
Lihavainen (2012)	Secondary	●			●	●	●		●						N
Markle-Reid (2010)	Secondary	●							●	●	●				CM
Metzelthin (2015)	Primary		●	●					●	●					APN
Parsons (2013)	Secondary								●	●	●	●			N
Ploeg (2010)	Primary		●	●					●		●		●		APN
Spoorenberg (2018)b	Primary + Secondary		●	●	●			●							APN or SW
Suijker (2016)	Primary		●	●					●	●					APN
Szanton (2011)	Secondary	●								●					N and OT
Tuntland (2015)	Primary								●	●					PT and OT
Zimmer (1985)	Primary		●	●				●							APN or GP or SW
Dolovich (2019)	Primary	Team composition not reported													GP
Ford (2019)	Primary and Secondary	Team composition not reported													GP
Markle-Reid (2006)	Secondary	Team composition not reported													N
Shapiro (2002)	Secondary	Team composition not reported													CM

CM, care manager (profession not reported); G, geriatrician; N, nurse; OT: occupational therapist; Psy, psychologist; PT, physiotherapist; SW, social worker.

^aOnly applicable to Screening + Nurse-Led Care group.

^bOnly applicable to Complex Care Needs group and Frail group.

Three (6%) RCTs reported the use of additional components to enhance usual care, including the provision of health educational materials [19–21] and standard needs assessment [22].

Risk of bias assessment

Risk of bias assessment is shown in Table S4. Overall, 11 were low risk of bias, 25 moderate and 14 at high risk. Twenty-seven studies had moderate risk of bias for not reporting details on randomisation and/or allocation sequence concealment, and one [23] was at high risk of bias for not concealing allocation sequence. Thirteen RCTs were at high risk of bias because they adopted per-protocol analysis for all outcomes, and three [24–26] had moderate risk of bias for not implementing the intention-to-treat analysis on all outcomes. Twenty-three studies had moderate risk of bias in the selection of reported results for not providing accessible study protocols.

Effects of interventions

Primary outcomes

Living at home

In 11 RCTs with 4,538 participants, interventions were significantly superior to usual care in increasing the likelihood of older adults living at home (RR 1.05; 95% confidence interval (CI) 1.00–1.10; P = 0.048; I² = 46%; GRADE moderate-quality evidence) (Table 4 and Figure S1). In the sensitivity analysis, there was no significant difference (P = 0.46) between studies at low/moderate versus high risk of bias. The leave-one-out analysis found that the significance of the pooled results was sensitive to seven studies (results became non-significant) [22, 23, 27–31] (Table S5). Little evidence of small-study effects was observed (Egger’s test: P = 0.21).

Table 4

Effect estimates and quality of evidence ratings for all outcomes

Dichotomous outcomes
Outcome	Number of RCTs (Number of participants)	Risk of bias	Inconsistency	Indirectness	Imprecision	Publication bias	Pooled result in RR (95% CI)	P	I2	Quality of evidence
Living at home	12 (4538)	No	Serious	No	No	Not detected	1.05 (1.00–1.10)	0.048	46%	|$\bigoplus $||$\bigoplus $||$\bigoplus $|◯ Moderate
Mortality	20 (9455)	No	No	No	No	Not detected	0.86 (0.77–0.96)	0.007	9%	|$\bigoplus $||$\bigoplus $||$\bigoplus $||$\bigoplus $| High
Hospitalisation	15 (6244)	No	Very serious	No	No	Not detected	0.93 (0.84–1.03)	0.19	59%	|$\bigoplus $||$\bigoplus $|◯◯ Low
Institutionalisation	15 (5231)	No	No	No	No	Not detected	0.89 (0.75–1.04)	0.14	23%	|$\bigoplus $||$\bigoplus $||$\bigoplus $||$\bigoplus $| High
Continuous outcomes
Outcome	Number of RCTs (Number of participants)	Risk of bias	Inconsistency	Indirectness	Imprecision	Publication bias	Pooled result in SMD (95% CI)	P	I2	Quality of evidence
QoL (overall)	9 (9460)	No	No	No	No	Not assessed	0.01 (−0.04 to 0.05)	0.72	0%	|$\bigoplus $||$\bigoplus $||$\bigoplus $||$\bigoplus $| High
QoL (physical component)	6 (5902)	No	Very serious	No	No	Not assessed	0.00 (−0.08 to 0.08)	0.97	54%	|$\bigoplus $||$\bigoplus $|◯◯ Low
QoL (mental component)	6 (5902)	No	Very serious	No	No	Not assessed	0.07 (−0.02 to 0.16)	0.11	59%	|$\bigoplus $||$\bigoplus $|◯◯ Low
Cognitive function	5 (2149)	No	No	No	No	Not assessed	0.12 (0.02–0.22)	0.02	0%	|$\bigoplus $||$\bigoplus $||$\bigoplus $||$\bigoplus $| High
Functional status (ADL)	6 (2476)	No	Serious	No	No	Not assessed	0.10 (0.00–0.20)	0.052	26%	|$\bigoplus $||$\bigoplus $||$\bigoplus $|◯ Moderate
Functional status (IADL)	4 (1687)	No	No	No	No	Not assessed	0.11 (0.01–0.21)	0.02	0%	|$\bigoplus $||$\bigoplus $||$\bigoplus $||$\bigoplus $| High
Functional status (Combined ADL and IADL)	5 (7751)	No	No	No	No	Not assessed	0.08 (0.03–0.13)	0.002	0%	|$\bigoplus $||$\bigoplus $||$\bigoplus $||$\bigoplus $| High

QoL (overall)

In nine RCTs with 9,460 participants, interventions made little or no difference to overall QoL in older adults (SMD 0.01; 95% CI –0.04 to 0.05; P = 0.72; I² = 0%; GRADE high-quality evidence) (Table 4 and Figure S2). Neither of the two unpooled RCTs reported significant results on the outcome (Table S2). In the leave-one-out analysis, the significance of the pooled results was not sensitive to any individual studies (Table S5).

QoL (physical component)

In six RCTs with 5,902 participants, interventions had little or no effect on the physical component of QoL (SMD 0.00; 95% CI –0.08 to 0.08; P = 0.97; I² = 54%; GRADE low-quality evidence) (Table 4 and Figure S3). Neither of the two unpooled RCTs reported significant impacts on the outcome (Table S2). No significant difference (P = 0.41) was identified between studies reporting change scores from the baseline versus studies reporting follow-up scores. The significance of the pooled results was not sensitive to any individual study (Table S5).

QoL (mental component)

In six RCTs with 5,902 participants, interventions had little or no effect on the mental component of QoL (SMD 0.07; 95% CI –0.02 to 0.16; P = 0.11; I² = 59%; GRADE low-quality evidence) (Table 4). The single unpooled RCT did not report significant results on the outcome (Table S2). There was a significant difference (P = 0.01) between studies (n = 3) reporting change scores from the baseline and the single study reporting scores at the end of follow-up (Figure S4), with the latter [32] having a positive result. The significance of the pooled results was sensitive to van Hout et al. [33] (results became significant) (Table S5).

Secondary outcomes

Mortality

In 20 RCTs with 9,455 participants, interventions reduced mortality in older adults (RR 0.86; 95% CI 0.77–0.96; P = 0.007; I² = 9%; GRADE high-quality evidence) (Table 4 and Figure S5). Only one [19, 20] of the six unpooled RCTs reported that the interventions were superior to usual care in reducing mortality at 12-month follow-up (Table S2). Little evidence of small-study effects was observed (Egger’s test: P = 0.10).

Hospitalisation

In 15 RCTs with 6,244 participants, interventions had little or no effect on hospitalisation (RR 0.93; 95% CI 0.84–1.03; P = 0.19; I² = 59%; GRADE low-quality evidence) (Table 4 and Figure S6). None of the five unpooled RCTs reported significant results for hospitalisation (Table S2). Little evidence of small-study effects was observed (Egger’s test: P = 0.56).

Institutionalisation

In 15 RCTs with 5,231 participants, interventions had little or no effect on institutionalisation (RR 0.89; 95% CI 0.75–1.04; P = 0.14; I² = 23%; GRADE high-quality evidence) (Table 4 and Figure S7). None of the four unpooled RCTs reported significant results for institutionalisation (Table S2). Little evidence of small-study effects was observed (Egger’s test: P = 0.21).

Cognitive function

In five RCTs with 2,149 participants, interventions were effective in improving cognitive function (SMD 0.12; 95% CI 0.02–0.22; P = 0.02; I² = 0%; GRADE high-quality evidence) (Table 4 and Figure S8). One [34] of the two unpooled RCTs reported that the interventions were superior to usual care in slowing down cognitive decline at 12-month follow-up (Table S2).

Functional status (ADLs, IADLs and combined ADLs/IADLs)

In six RCTs with 2,476 participants, interventions had little or no effect on ADLs (SMD 0.10; 95% CI 0.00–0.20; P = 0.052; I² = 26%; GRADE moderate-quality evidence) amongst older adults associated with complex interventions (Table 4 and Figure S9). However, one [35] of the two unpooled RCTs reported that the interventions were more effective than usual care in improving ADLs at 12-month follow-up (Table S2). Interventions had positive effects on IADLs (SMD 0.11; 95% CI 0.01–0.21; P = 0.02; I² = 0%; four RCTs with 1,687 participants; GRADE high-quality evidence) and combined ADLs/IADLs (SMD 0.08; 95% CI 0.03–0.13; P = 0.002; I² = 0%; five RCTs with 7,751 participants; GRADE high-quality evidence) (Table 4 and Figures S10 and S11). Although neither of the two unpooled RCTs measuring IADLs reported significant results, the unpooled RCT measuring combined ADLs/IADLs showed that the interventions were more effective than usual care at 12-month follow-up (Table S2). Amongst these three outcomes, no significant differences (P = 0.62; P = 0.74; P = 0.91) were identified between studies reporting change scores from the baseline versus studies reporting follow-up scores.

Functional status (physical mobility)

Four RCTs measured the change in physical mobility of 1,475 participants. Compared with usual care, interventions reduced self-reported difficulties in walking at 36-month follow-up [36], improved the Short Physical Performance Battery overall, balance and gait speed scores at 6-month follow-up [37] and increased the total hand grip strength at 6-month follow-up [38] (Table S2).

Subgroup analysis

Subgroup analyses found that the location of intervention delivered (P = 0.01), home/telephone follow-up (P = 0.03) and self-management (P = 0.03) modified the effect of community-based complex interventions on institutionalisation (Figures S12–S14). Home-based interventions were associated with a lower institutionalisation rate amongst older adults (RR 0.65; 95% CI 0.48–0.87; P = 0.004; I² = 0%). Interventions involving scheduled home/telephone follow-up (RR 0.75; 95% CI 0.60–0.93; P = 0.01; I² = 17%) or self-management (RR 0.58; 95% CI 0.38–0.88; P = 0.01; I² = 0%) also reduced institutionalisation rate amongst the population. However, the covariates in the analyses were unevenly distributed, meaning that the findings should be interpreted with caution [18]. No significant difference was identified in analyses of subgroups defined by length of follow-up, frailty, disability or functional decline of participants, or multidisciplinary care.

Discussion

Summary of findings

This systematic review found that holistic assessment and care plan development are the core components of >90% community-based complex interventions for improving independent living and QoL of older adults. Meta-analyses showed that interventions increased the likelihood of living at home (moderate-quality evidence) but had little to no effect on improving QoL (high-quality evidence for overall QoL; low-quality evidence for physical and mental components of QoL). Interventions also reduced mortality (high-quality evidence) and improved cognitive function (high-quality evidence), IADLs (high-quality evidence) and combined ADLs/IADLs (high-quality evidence). Although there was no impact on institutionalisation in the main analysis, subgroup analysis found significant reductions in institutionalisation for interventions delivered at home, with scheduled home/telephone follow-up, or with planned self-management.

Strengths and limitations

Strengths of this review include the performance of a comprehensive literature search in multiple databases and the adoption of the GRADE approach to evaluate quality of evidence. The review is also distinctive in including a wide range of trials with similar intervention components rather than searching based on the labels given to the intervention by researchers (e.g. CGA). Although the heterogeneity of interventions was a potential limitation, considerable heterogeneity within studies with the same label was also present [4, 8].

There are several further limitations. First, intervention components were variably and very likely incompletely reported by many studies. For this reason, we did not attempt network meta-analysis. Similarly, the limited number of studies and imbalanced covariates between subgroups meant that meta-regression could not be robustly applied [13]. Instead, we used subgroup analysis to explore potential effect modifiers of multifaceted interventions, but interpretation should be cautious given variable reporting, and some subgroup analyses were not possible (for example, further analysis of subgroups defined by multidisciplinary team make-up and the status or degree of frailty of the participants in the primary studies, because of lack of provision of this information in the original papers). Second, some studies did not report mortality, hospitalisation and institutionalisation in a way which could be pooled in meta-analyses, so could only be synthesised narratively. Third, mortality was not accounted for as a competing risk in individual study analyses [39]. Given the intervention impact on mortality, the interpretation of effects on hospitalisation and institutionalisation should be cautious. Fourth, given the variability in health and social care systems and infrastructures, the findings in this review may not be applicable in all settings. Finally, as only studies published in English were eligible for inclusion, the effect sizes in this review might have been overestimated or underestimated.

Comparisons with other reviews

Ellis et al. [5] found high-quality evidence that inpatient CGA increased the chance of living at home at discharge and decreased nursing home admissions but had no effect on mortality. Chen et al. [7] found that inpatient and community-based CGA improved the QoL of older adults, but the effects were not significant in the community-based subgroups. In reviews of interventions for community-dwelling older adults, Briggs et al. [6] found a decreased risk of hospital admissions but no effect on mortality and nursing home admissions, and Wong et al. [40] reported possible benefits on the mental component of QoL but not on overall and the physical component of QoL, or on ADL/IADL. The review [8] for the NICE multimorbidity guidelines mentioned that complex interventions had limited benefits in critical outcomes (e.g. mortality and QoL). Our study found evidence of benefits for some outcomes (living at home, mortality, cognitive function, IADLs and combined ADLs/IADLs) but not others (QoL, hospitalisation or institutionalisation). The more favourable findings may be because we included a larger range of complex interventions (with shared core components) rather than only including interventions with specific labels such as ‘CGA’ (which is not a homogeneous group either, given variable intervention components).

Implications for practice

We focused this review on complex interventions for community-dwelling older adults with similar components instead of relying on intervention labels, such as CGA. Two near ubiquitous components were identified: (i) holistic assessment (94% of trials) and (ii) care plan development (90% of trials), and these should therefore be considered as the cores for health services planning to implement such interventions. There was some evidence that scheduled home/telephone follow-up and self-management positively modified the effect of complex interventions on institutionalisation. There was no clear evidence that multidisciplinary care was beneficial, although its component was poorly reported by the included studies. Before implementing holistic assessment, organisers should ensure the trust between health and social care professionals carrying out the task and facilitate inter-professional communications [41]. The scope of assessment should also balance the needs of older adults with complicated problems and the limited assessment time [41]. However, specialist staff may not be necessary to in-home assessment for an ideal model of complex interventions [42, 43].

Implications for research

A weakness of the existing literature is poor reporting of both intervention components and ‘usual care’, including the lack of clarity about multidisciplinary team composition and the frequency and duration of intervention components. Detailed reporting using the Template for Intervention Description and Replication [44] or the Criteria for Reporting the Development and Evaluation of Complex Interventions in healthcare 2 [45] checklists would significantly improve interpretation and future evidence synthesis. Similarly, the trials examined in this review varied considerably in the outcomes measured, and how outcomes were measured. The use of a core outcome set with standardised instruments in future trials would ensure that future evidence is more comparable and easier to synthesise [46]. The core outcome set development should involve a panel of key stakeholders who will utilise, deliver and/or evaluate the complex interventions (i.e. researchers, clinicians, policy-makers, older adults and caregivers) in the form of Delphi surveys or semi-structured group discussions [46]. Outcomes for research on geriatric rehabilitation [47], older adults with frailty [48] and participants with multimorbidity [49], including health-related QoL, ADL/IADL and mental health, may also be adopted.

Conclusions

Holistic assessment and care plan development are the common components of complex interventions for improving independent living and QoL of community-dwelled older adults. Complex interventions increased the likelihood of living at home but had little to no effect on improving QoL. They reduced mortality and improved cognitive function, IADLs and combined ADLs/IADLs. Subgroup analyses suggested that complex interventions involving scheduled home/telephone follow-up or self-management might reduce institutionalisation rate amongst older adults; however, further evidence is needed to confirm such findings.

Declaration of Conflicts of Interest

None.

Declaration of Sources of Funding

The study was funded by Legal and General PLC (as part of their Corporate Social Responsibility programme, providing a research grant to establish the independent Advanced Care Research Centre at the University of Edinburgh). The funder had no role in conduct of the study, interpretation or the decision to submit for publication.

Data Availability Statement

All study data is provided in the paper and supplementary material.