Community-based management strategies for adults with multimorbidity: a systematic review of clinical and patient-centred outcomes

Abstract

Objectives

To assess the effectiveness of community-based interventions for adults with multimorbidity on clinical and patient-centred outcomes, and to examine contextual factors influencing their impact in primary care and community settings.

Design

Systematic review.

Data sources

PubMed/MEDLINE, Embase, Web of Science, Cochrane CENTRAL, China National Knowledge Infrastructure, WanFang Data and SinoMed were searched up to May 2025. Grey literature and trial registries were also searched.

Eligibility criteria

We included randomised controlled trials (RCTs), quasi-experimental studies and comparative observational studies involving adults (≥18 years) with multimorbidity, defined as ≥2 chronic conditions including at least one of hypertension, diabetes or dyslipidaemia. Interventions had to be delivered in primary care or community settings and report at least one clinical or patient-centred outcome.

Data extraction and synthesis

Two reviewers independently screened studies, extracted data and assessed risk of bias using the Cochrane RoB 2.0 and ROBINS-I tools. Due to heterogeneity in interventions and outcomes, results were synthesised narratively following the Synthesis Without Meta-analysis guidelines. Certainty of evidence was assessed using the Grading of Recommendations Assessment, Development and Evaluation approach.

Results

25 studies were included, comprising 19 RCTs and 6 quasi-experimental or observational designs. Interventions included nurse-led or multidisciplinary care (n=13), integrated or collaborative care (n=8) and digital health models (n=5). Clinical outcomes such as blood pressure, HbA1c or lipids were assessed in 10 studies, with 7 reporting significant improvements, 6 identifying subgroup-specific benefits and 3 reporting mixed or null effects. Patient-centred outcomes were reported in all studies; quality of life improved in 10 of 13 studies, self-management in 6 of 9 and healthcare use was reduced in 7 of 11.

Conclusions

Community-based interventions for multimorbidity consistently improve patient-centred outcomes, while clinical effects are more variable and context-dependent. Tailored implementation for high-risk groups and attention to local delivery models may enhance effectiveness. Further research is needed to support equity-focused, long-term implementation.

PROSPERO registration number

CRD420251159790.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Multimorbidity is increasingly common and associated with high health system burden, yet care models often remain fragmented and single-disease focused. Previous reviews have shown modest improvements in patient-centred outcomes but inconsistent effects on clinical measures. Many interventions have not been evaluated in real-world or underrepresented settings.

WHAT THIS STUDY ADDS

• This review synthesises recent evidence from 25 studies, including pragmatic trials and diverse care models, highlighting consistent benefits of community-based interventions on patient-centred outcomes and more variable effects on clinical outcomes. The findings identify contextual factors and subgroups that modify effectiveness, such as baseline risk and socioeconomic disadvantage.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• The study supports prioritising patient-centred, integrated care models within primary care systems and tailoring interventions to high-risk populations. It also highlights the need for future research to focus on context-sensitive implementation, longer-term outcomes and equity-focused targeting strategies.

Introduction

Multimorbidity is an increasingly common global phenomenon, placing a critical burden on health systems.1,3 It affects more than one-third of adults worldwide, with projections showing a steep increase in the future.¹ Its prevalence is particularly high in primary care settings and among ageing populations. Multimorbidity is associated with increased mortality, diminished quality of life and accelerated functional decline, with the greatest burden borne by individuals with both mental and physical conditions.^{4 5} Patients with multimorbidity account for the majority of primary care consultations and disproportionately contribute to emergency admissions and prolonged hospital stays, driving escalating healthcare costs.^{6 7} The problem is socially patterned, occurring 10–15 years earlier and with greater severity in socioeconomically deprived groups,⁸ and affects more than 90% of older adults with conditions such as diabetes or stroke.⁹

Current models of care are poorly equipped to address this complexity. Most health systems remain organised around single diseases, leading to fragmented and inefficient care.^{10 11} Clinical guidelines rarely address multimorbidity and often offer contradictory recommendations, resulting in inappropriate polypharmacy, excessive treatment burden and heightened risk of adverse outcomes.12,14 Despite the scale of the challenge, patients with multimorbidity remain under-represented in clinical research, leaving substantial evidence gaps. Addressing this problem requires innovative, patient-centred, community-based strategies that integrate care across conditions, support self-management and are adaptable to diverse health system contexts.

Despite growing recognition of multimorbidity as a priority for health systems, there remains limited evidence on how best to organise care for these patients in community and primary care settings, where the burden is greatest.15,17 A range of strategies have been proposed, including nurse-led or pharmacist-led programmes, collaborative and integrated care models, digital health interventions, home-based monitoring and structured self-management support.^{15 16 18} While some studies have shown consistent improvements in patient-centred outcomes, such as quality of life and self-management, their effectiveness across clinical and patient-centred outcomes remains uncertain.^{15 16 18} This review builds on previous syntheses by incorporating a broader set of intervention types and study designs, including recent trials conducted in low-income and middle-income countries and digitally enabled settings. We explicitly examined both clinical and patient-centred outcomes,^{16 19} explored contextual factors shaping intervention impact and highlighted evidence from pragmatic and real-world evaluations, thereby extending the current evidence base on community-based care for multimorbidity.

Methods

Eligibility criteria

This systematic review included studies involving adults (aged ≥18 years) with multimorbidity, defined as the presence of two or more chronic conditions, with at least one being hypertension, diabetes or dyslipidaemia. Studies were eligible if they evaluated community-based management strategies implemented in primary care or community settings. Eligible interventions included team-based care, nurse-led or pharmacist-led programmes, health education initiatives, digital health tools, home monitoring, medication adherence support or integrated care models. Comparators included usual care or other active interventions. Studies were included if they reported at least one clinical outcome, such as blood pressure, HbA1c, fasting glucose, lipid levels or target achievement rates or one patient-centred outcome, including medication adherence, quality of life, healthcare utilisation, cardiovascular events, mortality or cost-effectiveness. Eligible study designs comprised randomised controlled trials (RCTs), quasi-experimental studies and real-world observational studies, including cohort studies, before–after studies with comparators and interrupted time series. Studies were excluded if they were conducted exclusively in inpatient settings, focused solely on surgical or device-based interventions, employed qualitative or cross-sectional methodologies or lacked sufficient outcome data for extraction and synthesis. This included studies that lacked numerical outcome data (eg, means, proportions or effect sizes), reported results only in descriptive text without group-specific values or did not present outcomes attributable to adults with multimorbidity.

Information sources and search strategy

A comprehensive literature search was conducted in May 2025 across the following electronic databases: PubMed/MEDLINE, Embase, Web of Science, Cochrane CENTRAL, China National Knowledge Infrastructure, WanFang Data and SinoMed. To identify unpublished and ongoing studies, we searched clinical trial registries and grey literature sources including ClinicalTrials.gov, the WHO International Clinical Trials Registry Platform and the Chinese Clinical Trial Registry.

The search strategy combined Medical Subject Headings and free-text terms related to (i) care setting (“community care,” “primary care,” “ambulatory care”), (ii) population (“multimorbidity,” “multiple chronic conditions,” “chronic disease,” “hypertension,” “diabetes,” “dyslipidemia”), (iii) intervention strategies (“care management,” “team-based care,” “collaborative care,” “nurse-led,” “pharmacist-led,” “self-management support,” “digital health,” “telemedicine,” “home monitoring,” “integrated care”), and (iv) study design (“randomized controlled trial,” “cluster randomized trial,” “pragmatic trial,” “quasi-experimental,” “cohort,” “observational,” “real-world evidence”). Boolean operators (AND/OR), proximity operators and truncation were used as appropriate for each database.

No restrictions on publication year were applied. Studies published in English or Chinese were eligible. Reference lists of relevant reviews and included articles were hand-searched to identify additional studies. Where available, trial protocols and registry entries were screened to verify reported outcomes and identify selective reporting. The complete database-specific search strings are provided in online supplemental material 1, developed iteratively with input from a medical librarian to ensure sensitivity and comprehensiveness.

Selection process

Two reviewers (LC and SX) independently screened the titles and abstracts of all identified articles for relevance, applying the prespecified inclusion and exclusion criteria. Full texts of potentially eligible studies were retrieved and evaluated independently by the same reviewers. Any disagreements were resolved through discussion or adjudication by a third reviewer (LX). Automation tools were employed only for initial deduplication; no machine learning tools were used for inclusion decisions.

Data collection and extraction

Data extraction was performed by one reviewer (LC) using a standardised form and verified by a second reviewer (SX). Any discrepancies were resolved through discussion or adjudicated by a third reviewer (LX). The following data were extracted: publication year, country, sample size, study design, intervention characteristics (including duration, frequency and delivery mode), comparator details, clinical and patient-centred outcomes and main findings. Outcomes were categorised by time points where possible (≤6 months, 6–12 months, >12 months). Effect sizes, such as ORs, HRs or risk ratios with 95% CIs, were recorded where available. Missing data or unclear reporting were noted, and assumptions were made explicitly where necessary.

Risk-of-bias assessment

For RCTs, we used the Cochrane Risk of Bias 2.0 tool, assessing five domains and assigning judgements of low risk, some concerns or high risk per domain. For non-randomised studies, we applied the Risk Of Bias In Non-randomized Studies—of Interventions (ROBINS-I) tool using the same three-level classification.²⁰ To summarise overall risk, we assigned a numerical score (0–5), with 1 point given for each domain rated as ‘some concerns’ or ‘high risk’. Studies with total scores of 0–1 were considered low risk overall, 2–3 moderate risk and 4–5 high risk. Two reviewers independently assessed the risk of bias (LC and SX), with disagreements resolved by consensus and the third reviewer (LX).

Effect measures

For each included study, the effect measures used to report outcomes were recorded. These included risk ratios, ORs, HRs and mean differences, depending on the nature of the data and outcome type. Where available, 95% CIs were extracted to assess the precision of the reported effects.

Synthesis methods

Given the anticipated heterogeneity in study designs, interventions and outcome measures, a meta-analysis was not conducted. Instead, a narrative synthesis approach was employed in accordance with the Synthesis Without Meta-analysis guidelines.²¹ Studies were grouped by intervention type and outcome domain. Key characteristics and results of the included studies were summarised in tabular and textual form. Data transformations or conversions were not performed. Visual displays were used to present study characteristics and outcome summaries. Sensitivity analyses were undertaken to assess the robustness of findings where study-level variability permitted.

Reporting bias assessment

To assess the risk of reporting bias, protocols and trial registry entries (when available) were compared with published reports to evaluate selective reporting of outcomes. The presence of reporting bias was assessed qualitatively, considering discrepancies between planned and reported outcomes. Reporting bias was categorised qualitatively as low, some concerns or high, based on the extent of selective outcome reporting or discrepancies between planned and reported outcomes.

Certainty of evidence

The certainty of evidence for the main outcomes was assessed using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach, taking into account risk of bias, inconsistency, indirectness, imprecision and potential publication bias. Certainty ratings were assigned as high, moderate, low or very low, in accordance with standard GRADE guidance.

Protocol registration and reporting standards

This review was registered prospectively in the PROSPERO international prospective register of systematic reviews (registration ID: CRD420251159790) and conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines.

Patient and public involvement

None

Results

Study characteristics

A total of 25 studies were included in this systematic review (table 1; online supplemental figure 1).22,46 The majority employed RCT designs (n=20), including pragmatic, open-label, single-blind, stratified, non-inferiority, superiority, assessor-blinded and secondary analysis approaches. Four of these were cluster RCTs. Three studies adopted quasi-experimental designs, one used a cluster-controlled design, two used a controlled trial and one used a trials-within-cohorts design⁴⁰ (table 1).

Table 1. Characteristics of the included studies.

Study	Study design	Population	Intervention type	Primary outcomes
Miklavcic et al35	Pragmatic RCT, multisite	Older adults (≥65 years), T2DM +≥2 comorbidities	Nurse-led, in-home and group visits	Physical functioning (SF-12 PCS)
Fortin et al26	Pragmatic RCT, mixed-methods	Adults 18–80 years, ≥3 chronic conditions	Interdisciplinary, motivational self-management	Self-management, self-efficacy
Ose et al37	RCT, open-label	T2DM multimorbid patients	Care management (IT-based, nurse-led)	Self-care behaviour (SDSCA-G)
Contant et al23	Pragmatic RCT	Adults 18–75 years, ≥3 chronic conditions	Multidisciplinary self-management	Self-management (heiQ)
Majert et al31	RCT, open-label, pilot	Older adults (≥65 years), multimorbidity/polypharmacy	Home-based antihypertensive titration (remote)	Change in home BP
Li et al29	RCT	Adults 18–75 years, ≥1 chronic condition, polypharmacy	Digital adherence app (Perx)	Medication adherence
Steinman et al42	Controlled trial	Adults 45–94 years, ≥3 chronic conditions	Nurse-based, guided care	Medication regimen changes
Lin et al30	RCT	Depressed patients w/ diabetes or CHD	Collaborative care	HbA1c, SBP, LDL, depression
Mercer et al34	Cluster RCT	Adults 30–65 years, multimorbidity in deprived areas	Whole-system primary care (CARE Plus)	QoL (EQ-5D-5L), well-being
Fisher et al25	Pragmatic RCT	Older adults (≥65 years), ≥3 chronic conditions	Interprofessional home-based care	Physical functioning (SF-12 PCS)
Mateo-Abad et al33	Quasi-experimental	Older adults (≥65 years), ≥2 chronic conditions	Integrated care (CareWell+ICT)	Health resource use, effectiveness
Yang et al 46	RCT, single-blind	Older adults w/ multimorbidity	Nurse-led medication self-management	Medication adherence
Markle-Reid et al32	Pragmatic RCT	Older adults (≥55 years) w/ stroke+multimorbidity	Transitional care (virtual, interprofessional)	Hospital readmission
Stewart et al, 202143	Pragmatic RCT, mixed-methods	Adults 18–80 years, ≥3 chronic conditions	Multi-provider case conference+follow-up	Self-management, self-efficacy
Netti Tiozzo et al44	Observational, matched	Elderly CHF+multimorbidity	Nurse-GP care management	Hospitalisation, ER visits
Lee et al28	RCT, secondary analysis	Older adults w/ multimorbidity, dementia syndrome	Multidomain (exercise, cognitive, diet)	Cognitive function, frailty, QoL
Skou et al41	RCT, assessor-blinded	Adults w/ ≥2 LTCs + activity limitation	Exercise therapy+self-management	QoL (EQ-5D-5L)
Wang et al45	Cluster RCT	Adults w/ T2DM+depression	Collaborative integrated care (diabetes+depression)	Cost-effectiveness (QALYs, depression-free days)
John et al27	Cohort, controlled	Hypertensive patients+comorbidities	Patient-centred medical home (WellNet)	Blood pressure outcomes
Otieno et al38	Quasi-experimental	Hypertensive patients, some w/ multimorbidity	Home-based self-care+support groups	BP control
Panagioti et al40	Trials within cohorts	Older adults, ≥2 LTCs	Telephone health coaching	Patient activation, QoL, depression
del Cura-González et al 2021	Cluster RCT	Older adults 65–74 years, multimorbidity+polypharmacy	MULTIPAP (Ariadne principles)	Medication appropriateness (MAI)
Camacho et al22	Cluster RCT	Adults w/ depression+diabetes/CHD	Collaborative care	Depression, QALYs
Oksman et al36	RCT	Adults w/ T2DM, CAD, CHF	Tele-based health coaching	QoL (15D), cost-effectiveness
Øverås et al39	RCT, single-blind	Adults w/ LBP±multimorbidity	AI-based self-management (selfBACK)	LBP disability (Roland-Morris)

BP, blood pressure; CHD, coronary heart disease; CHF, congestive heart failure; EQ-5D-5L, EuroQol 5-Dimension 5-Level instrument; ER, emergency room; GP, general practitioner; HbA1c, glycated haemoglobin; LBP, low back pain; LDL, low-density lipoprotein cholesterol; LTCs, long-term conditions; MAI, Medication Appropriateness Index; QALYs, quality-adjusted life years; QoL, quality of life; RCT, randomised controlled trial; SBP, systolic blood pressure; SDSCA-G, Summary of Diabetes Self-Care Activities (German version); SF-12 PCS, 12-Item Short Form Health Survey physical component summary; T2DM, type 2 diabetes mellitus.

10 studies included older adults (≥55 or ≥65 years),2224 25 27 28 31,33 35 40 while all 25 studies included broader adult populations ranging from 18 to 94 years. All studies explicitly enrolled patients with multimorbidity (defined as two or more chronic conditions), and three focused on populations with polypharmacy. Specific subgroups included type 2 diabetes mellitus with multimorbidity (n=2),^{35 37} type 2 diabetes mellitus with comorbid depression (n=3),^{22 30 45} coronary heart disease or cardiovascular disease (n=5).^{22 28 30 36 44} Other targeted populations included those with hypertension and comorbidities, stroke and multimorbidity, congestive heart failure and multimorbidity, low back pain with or without multimorbidity and high-risk multimorbid adults.

Interventions

Intervention strategies were diverse and often complex. Nurse-led, interprofessional or multidisciplinary approaches were evaluated in 14 studies.2223 25,27 32 Collaborative or integrated care models were tested in 8 studies,^{22 27 28 30 33 37 43 45} while 15 included structured self-management support interventions.2325,27 29 32 12 studies used digital or technology-based interventions, including mobile health, telehealth, digital adherence applications and artificial intelligence–based tools.2931,33 36 37 40 41 43 45 46 Three studies investigated exercise-based or multidomain programmes^{28 35 41} and four evaluated medication-related strategies such as medication management, review or deprescribing.^{24 25 31 42} Other approaches included care management or coordination, transitional care, health coaching, home-based interventions, patient-centred medical home models, support groups, multiprovider case conferences, whole-system or primary care reforms and participatory care planning.

Outcomes

Primary outcomes were heterogeneous but spanned both clinical and patient-centred domains. Quality of life and well-being were measured in 17 studies,2223 25 26 28 30,37 40 41 43 45 physical functioning or exercise capacity in 14,2325 26 28 29 31,33 35 37 41 43 depression or mental functioning in 14,2223 25 26 28 30,36 40 41 self-management or self-efficacy in 15,23,2629 30 32 34 35 37 40 41 43 45 46 medication adherence or appropriateness in 9,2429,31 35 40 42 45 46 blood pressure outcomes in 62729,31 36 41 and hospitalisation or emergency visits in 10.2224 25 31,34 36 40 44 Other outcomes included health service use or cost-effectiveness (n=15), cognitive function or frailty (n=3), disease-specific markers such as haemoglobin A1c, systolic blood pressure and low-density lipoprotein cholesterol (n=8), and single reports of composite outcomes, therapy use, care quality and low back pain disability. No included study was missing primary outcome reporting.

Clinical outcomes

Table 2 shows 12 of the 25 included studies reporting clinical outcomes. Among these, six reported blood pressure outcomes,2729,31 36 41 five haemoglobin A1c,^{29 30 33 37 45} three lipid parameters28,30 and seven other measures such as body mass index, glucose, composite cardiovascular events, peak oxygen uptake or polypharmacy.^{24 28 29 31 33 42 45} Of these 12, 7 demonstrated statistically significant improvements in at least one clinical endpoint, 2 showed significance only in a subgroup with poor baseline control, 1 reported borderline or non-significant effects and 2 showed no effect.

Table 2. Clinical outcomes of community-based management strategies.

Study	Intervention	Clinical measures	Effect size	Significance
Miklavcic et al35	Nurse-led, group-based	No BP/HbA1c data	No effect	Not significant
Fortin et al26	Interdisciplinary, motivational	Self-management (Health Education Impact Questionnaire) and self-efficacy (Self-Efficacy for Managing Chronic Diseases)	Neutral effect	Not significant
Ose et al37	Care management	HbA1c 7.1±1.2 vs 7.3±1.2; SBP 135.5 vs 133.9	No significant change	Not significant
Contant et al23	Multidisciplinary self-management	Self-management (Health Education Impact Questionnaire)	OR from 1.96 to 2.91	Significant in six of eight domains
Majert et al31	Home-based antihypertensive titration	SBP: 134.5→122.1 vs 134.8→132.9	−10.7 mm Hg	Significant
Li et al29	Digital adherence application	HbA1c≤6.5% higher at 9–12 months; LDL-C lower at 3 months	7% higher adherence at 12 months	Significant
Steinman et al42	Nurse-based, guided care	Number of medication regimen changes at 9 months: 4.04 vs 3.62	+0.55 medication changes	Significant
Lin et al30	Collaborative care	HbA1c, SBP, LDL improved in poor baseline control	Sustained effect	Significant (subgroup)
Mercer et al34	Whole-system primary care	Quality of life (EQ-5D-5L utility scores) and well-being (W-BQ12)	Effect size of 0.33 for the negative well-being domain of the W-BQ12 at 12 months The effect size for quality of life (EQ-5D-5L) measured as area under the curve was 0.36	Significant
Fisher et al25	Interprofessional, home-based	Physical Component Summary score from the SF-12 health survey	−4.94 points	Not significant
Mateo-Abad et al33	Integrated care (CareWell)	SBP 132.3→127.5; HbA1c 6.7→6.6	−4.7 mm Hg, HbA1c −0.33	Borderline/NS
Yang et al46	Nurse-led medication self-management	Medication adherence (MARS-5) at post-intervention and 3 month follow-up	β=1.63 (post-intervention)	Significant
Markle-Reid et al32	Transitional care	Physical functioning (SF-12 Physical Component Summary), stroke self-management (SSSMQ), patient experience (P3CEQ) and hospital readmission at 6 months	Physical functioning mean difference: 5.10; self-management mean difference: 6.00	Significant
Stewart et al 202043	Multiprovider case conference	Self-management (Health Education Impact Questionnaire), self-efficacy (SEM) and mental health status (VR12 MCS) at 4 months	Neutral effect on primary outcomes.	Not significant
Netti Tiozzo et al44	Nurse-GP care management	Hospital admissions and ER visits over 1 year	−39% in hospitalisation rates; −33% in ER visits	Significant for hospitalisations (p=0.025); borderline for ER visits (p=0.06)
Lee et al28	Multidomain	Global cognitive performance (MoCA score) among older adults with multimorbidity and physio-cognitive decline syndrome	+1.1 points	Significant
Skou et al41	Exercise and self-management	Health-related quality of life (EQ-5D-5L) at 12 months	0.064 points	Significant
Wang et al45	Integrated care (diabetes+depression)	Composite outcome: ≥50% reduction in depressive symptoms (SCL-20) and≥0.5% reduction in HbA1c at 12 months	Risk difference for depression: 31.03%; risk difference for HbA1c: 19.16%	Significant
John et al27	Patient-centred medical home	SBP −3.4 mmHg	−3.4 mm Hg	Significant
Otieno et al38	Home-based self-care and support	SBP 136.5→133.0 vs 90→87	−5.4 mm Hg	Significant
Panagioti et al40	Telephone health coaching	Patient activation (PAM) and quality of life (WHOQOL-BREF)	No significant benefit on patient-reported outcomes; likely cost-effective	Not significant
del Cura-González et al, 2021	MULTIPAP (Ariadne)	Medication appropriateness (MAI score) at 6 and 12 months	Mean difference: −2.42 (6 months), −3.40 (12 months)	Significant
Camacho et al22	Collaborative care	Depression score (SCL-D13) at 24 months	−0.27 points	Significant
Oksman et al36	Tele-based health coaching	Cost-effectiveness (ICER per QALY)	T2D: €20,000/QALY; CAD: €40,278/QALY; CHF: negative effect	Cost-effective for T2D and CAD groups at certain willingness-to-pay thresholds
Øverås et al39	AI-based self-management app	LBP-related disability (RMDQ)	No modification of effect by multimorbidity or co-occurring musculoskeletal pain	Not significant

AI, artificial intelligence; BP/HbA1c, blood pressure and glycated haemoglobin; CAD, coronary artery disease; CER, incremental cost-effectiveness ratio; CHF, congestive heart failure; EQ-5D-5L, EuroQol 5-Dimension 5-Level instrument; ER, emergency room; GP, general practitioner; LBP, low back pain; LDL, low-density lipoprotein; LDL-C, low-density lipoprotein cholesterol; MAI, Medication Appropriateness Index; MARS-5, 5-item Medication Adherence Report Scale; MoCA, Montreal Cognitive Assessment; NS, not significant; PAM, Patient Activation Measure; P3CEQ, Person-Centred Coordinated Care Experience Questionnaire; QALY, quality-adjusted life year; RMDQ, Roland-Morris Disability Questionnaire; SBP, systolic blood pressure; SEM, self-efficacy measure; SF-12, 12-Item Short Form Health Survey; SSSMQ, Stroke Self-Management Questionnaire; T2D, type 2 diabetes; VR12 MCS, Veterans RAND 12-Item Health Survey mental component summary; WHOQOL-BREF, World Health Organization Quality of Life-BREF.

Integrated care interventions were most common among studies reporting clinical outcomes (n=5).^{22 30 33 44 45} Digital health interventions were tested in four of these studies,^{22 30 33 45} with three showing significant benefit.^{22 30 45} Four studies evaluated home-based interventions,^{25 31 35 44} of which one reported significant reductions in systolic blood pressure.³¹ Collaborative care models were tested in four studies,^{22 26 30 45} all of which demonstrated subgroup-specific benefit.^{22 26 30 45} Most nurse-led and multidisciplinary programmes did not report direct clinical outcome measures.

Patient-centred outcomes

Patient-centred outcomes were more frequently reported (table 3). Nurse-led interventions (n=6),^{25 35 36 42 44 46} collaborative care models (n=5),^{22 26 30 43 45} multidisciplinary or whole-system interventions (n=14),2224,26 30 32 integrated care (n=8)2226 30 32 33 43,45 and digital health (n=7)2729 31,33 44 45 were the most frequently assessed intervention categories. Home-based titration (n=1),³¹ care management (n=4),^{27 37 42 44} medication management (n=3),^{24 42 46} exercise-based approaches (n=2)^{27 41} and transitional care (n=2)^{32 44} were also represented. Other single-study interventions included patient-centred medical home models, case conferences, support groups and health coaching.

Table 3. Patient-centred outcomes of community-based management strategies.

Study	Intervention	Patient-Centred measures	Effect direction	Generalisability
Miklavcic et al35	Nurse-led, group-based	Quality of life, mental health, cost	Neutral	Older, high comorbidity, Canada
Fortin et al26	Interdisciplinary	Self-management, self-efficacy, lifestyle	Mixed (neutral quantitative, positive qualitative)	Canada, family medicine groups
Ose et al37	Care management	Self-care behaviour	Trend positive	Germany, high baseline
Contant et al23	Multidisciplinary	Self-management (Health Education Impact Questionnaire)	Improved	Canada, low SES impact
Majert et al31	Home-based titration	Adverse events, safety	No increase	Older, UK, remote
Li et al29	Digital adherence application	Adherence, clinical outcomes	Improved	Australia, polypharmacy
Steinman et al42	Nurse-based care	Medication changes	More attentive	Israel, high-risk
Lin et al30	Collaborative care	Depression, sustained benefit	Improved in poor control	USA, stratified
Mercer et al34	Whole-system care	Quality of life, well-being, cost-effectiveness	Improved	Deprived, Scotland
Fisher et al25	Interprofessional	Quality of life, self-efficacy, cost	Neutral	Canada, older adults
Mateo-Abad et al33	Integrated care	Satisfaction, empowerment	Improved	Spain, older adults
Yang et al46	Nurse-led medication self-management	Adherence, knowledge, beliefs	Improved (short-term)	China, older adults
Markle-Reid et al32	Transitional care	Physical function, self-management, experience	Improved	Canada, stroke, older
Stewart et al, 202143	Case conference	Mental health, experience	Improved in subgroup	Canada, income effect
Netti Tiozzo et al44	Nurse-GP care	Hospitalisation, ER, PACIC	Improved	Italy, elderly CHF
Lee et al28	Multidomain	Cognitive, frailty, QoL	Improved	Taiwan, dementia
Skou et al41	Exercise and self-management	Health-related QoL, self-rated health	Improved	Europe, long-term conditions
John et al27	Patient-centred home	Blood pressure control	Improved	Australia, large sample
Otieno et al 2023	Self-care and support	Blood pressure control, group effect	Improved, attenuated by multimorbidity	Kenya, LMIC
Panagioti et al40	Health coaching	Activation, quality of life, cost	Neutral, cost-effective	UK, older adults
del Cura-González et al 2021	MULTIPAP	Medication appropriateness	Improved	Spain, polypharmacy
Camacho et al22	Collaborative care	Depression, QALYs, cost	Improved	UK, mental-physical
Oksman et al36	Tele-coaching	Quality of life, cost-effectiveness	Improved	Finland, subgroup effect
Øverås et al39	AI self-management	Low back pain disability	Neutral	Norway/Denmark, LBP+multimorbidity

CHF, congestive heart failure; ER, emergency room; GP, general practitioner; LBP, low back pain; LMIC, low- and middle-income country; PACIC, Patient Assessment of Chronic Illness Care; QALY, quality-adjusted life year; QoL, quality of life; SES, socioeconomic status.

The most common patient-centred outcomes were quality of life, satisfaction or patient experience, self-management or self-care, depression or mental health, physical function or exercise capacity, cost-effectiveness, cognitive function, adherence, empowerment and hospitalisations. Other outcomes included medication appropriateness or regimen changes, knowledge and beliefs, blood pressure control and quality-adjusted life years.

In terms of direction of effect, improved patient-centred outcomes were observed in 13 studies,2226 29 30 32,36 41 neutral effects in 1,²⁶ mixed effects in 92430,34 40 41 43 and no difference in 2 studies.^{25 35} Subgroup benefits were reported in seven studies (eg, greater improvement in patients with poor baseline control,³⁰ those with higher income³⁴ or specific disease subgroups).^{28 29 38 45} Positive short-term effects were observed in one study,⁴⁶ while intervention benefits were attenuated by multimorbidity in another.³⁸ Importantly, no study reported a negative effect direction.

Intervention-specific patterns and contextual factors

Multidisciplinary, collaborative and integrated care interventions were frequently associated with improvements, though these were most consistently seen in patient-centred outcomes rather than clinical markers.^{22 43} For example, studies reported significant benefits in depression symptoms,²² patient experience,^{32 33} self-management^{23 32} and shifts in health service use towards primary care and away from emergency rooms.³³ However, effects on clinical outcomes such as physical functioning or disease-specific markers were often neutral or mixed.^{26 32 33 35} Many of these interventions explicitly incorporated self-management support as a key component.^{23 32 34 45}

Digital and technology-based interventions demonstrated feasibility, but their effectiveness was highly dependent on context and engagement.⁴⁰ A smartphone app improved medication adherence and some clinical outcomes,²⁹ and an AI-based app for low back pain was effective regardless of multimorbidity status.³⁹ Remote medication management also successfully lowered blood pressure.³¹ In contrast, a telephone health coaching intervention suffered from low uptake (41% consent rate), which likely contributed to a lack of significant improvement in patient-reported outcomes, even though an economic analysis suggested it was likely cost-effective.⁴⁰

Exercise and self-management support interventions were shown to improve health-related quality of life,⁴¹ physical functioning³² and health behaviours like diet and physical activity.²⁶ A multidomain programme that included exercise, cognitive training and diet education improved cognitive function and quality of life for older adults with physio-cognitive decline syndrome.²⁸ Similarly, medication management strategies effectively improved prescribing appropriateness,²⁴ increased the rate of ‘fine-tuning’ adjustments to medication regimens⁴² and improved short-term medication adherence, although long-term effects were not always sustained.⁴⁶

The effectiveness of interventions was strongly influenced by population and healthcare context. Several studies reported greater benefit in patients with poor baseline control of their conditions^{27 30} or in high-risk or socioeconomically deprived populations.^{34 45} Conversely, some studies with neutral findings noted that their patient populations already had high levels of self-care or good clinical control at baseline, leaving little room for improvement.^{26 37} Contextual factors such as the quality of usual care,³⁵ the COVID-19 pandemic³² and the fidelity of intervention delivery frequently shaped the magnitude of the effect.^{26 37 43} Real-world pragmatic trials contributed to generalisability but also highlighted implementation challenges, such as low uptake or variations in delivery across sites, that can dilute an intervention’s impact.^{35 37 40 43}

Risk-of-bias assessment

Risk-of-bias assessment revealed variability across the included studies (tables 4 and 5). Among the RCTs evaluated using the Cochrane RoB 2.0 tool (table 4), most domains were judged as low risk, particularly for deviations from intended interventions and selection of reported results. However, several trials had some concerns in randomisation processes, outcome measurement or missing data, and three trials were judged to be at high risk of bias due to substantial missing outcome data. For the non-randomised studies assessed with ROBINS-I (table 5), the overall risk of bias ranged from moderate to serious, with the main concerns arising from confounding, participant selection and missing data. In particular, one study was judged to be at serious risk due to substantial missing data and another due to outcome measurement limitations.

Table 4. Risk-of-bias assessment for the included randomised trials using Cochrane RoB 2.0.

Study	D1. Randomisation process	D2. Deviations from intervention	D3. Missing outcome data	D4. Outcome measurement	D5. Selection of reported result	Overall Risk-of-bias score (0–5)
Camacho et al22	Some concerns	Low risk	Some concerns	Low risk	Low risk	2
Contant et al23	Some concerns	Low risk	Low risk	Some concerns	Low risk	2
del Cura-González et al24	Low risk	Low risk	Low risk	Low risk	Low risk	0
Fortin et al26	Low risk	Low risk	Low risk	Some concerns	Low risk	2
Wang et al45	Low risk	Low risk	Low risk	Low risk	Low risk	0
Fisher et al25	Low risk	Low risk	High risk	Low risk	Low risk	5
Li et al29	Low risk	Low risk	High risk	High risk	Low risk	5
Majert et al31	Low risk	Low risk	Low risk	Low risk	Low risk	0
Markle-Reid et al32	Low risk	Some concerns	Low risk	Low risk	Low risk	2
Mercer et al34	Low risk	Low risk	Low risk	Low risk	Low risk	0
Miklavcic et al35	Some concerns	Low risk	Low risk	Low risk	Low risk	2
Oksman et al36	Some concerns	Low risk	High risk	Some concerns	Low risk	5
Ose et al 201937	Low risk	Some concerns	Low risk	Some concerns	Low risk	2
Panagioti et al40	Low risk	Some concerns	Low risk	Low risk	Low risk	2
Skou et al41	Low risk	Low risk	Low risk	Low risk	Low risk	0
Stewart et al, 202143	Low risk	Low risk	Low risk	Some concerns	Low risk	2
Lee et al28	Low risk	Low risk	Low risk	Low risk	Low risk	0
Lin et al30	Low risk	Low risk	Low risk	Some concerns	Low risk	2
Yang et al46	Low risk	Low risk	Some concerns	Low risk	Low risk	2
Øverås et al39	Low risk	Low risk	Low risk	Low risk	Low risk	0

Table 5. Risk-of-bias assessment for non-randomised studies (ROBINS-I).

Study	Confounding	Selection of participants	Classification of interventions	Deviations from interventions	Missing data	Measurement of outcomes	Selection of reported result	Overall risk-of -bias score (0–5)
John et al27	Moderate risk	Moderate risk	Low risk	Low risk	Serious risk	Moderate risk	Low risk	5
Mateo-Abad et al33	Moderate risk	Moderate risk	Low risk	Moderate risk	Moderate risk	Moderate risk	Low risk	2
Otieno 202338	Low risk	Moderate risk	Low risk	Moderate risk	Moderate risk	Serious risk	Low risk	5
Steinman et al42	Moderate risk	Moderate risk	Low risk	Moderate risk	Moderate risk	Moderate risk	Low risk	2
Netti Tiozzo et al44	Moderate risk	Moderate risk	Low risk	Low risk	Low risk	Low risk	Low risk	2

Discussion

This systematic review synthesised evidence from 25 studies that evaluated a diverse range of community-based interventions for adults with multimorbidity and related conditions such as polypharmacy. The interventions were heterogeneous, encompassing integrated and collaborative care models, self-management support and health coaching, digital health technologies, medication management strategies, and exercise-based or multidomain programmes. Across these studies, improvements were more consistently reported in patient-centred outcomes, such as quality of life, self-management behaviours and patient experience, than in traditional clinical markers including HbA1c, blood pressure or lipid levels. Effects on clinical outcomes were mixed, with some interventions demonstrating clear benefits, particularly for patients with poor baseline control, while others reported neutral findings. Importantly, the effectiveness of interventions was frequently moderated by contextual and patient-level factors, including the healthcare system, quality of usual care, baseline characteristics and subgroup-specific needs. Our findings highlight both the potential and the limitations of community-based strategies in addressing the complex needs of multimorbid populations.

Compared with previous reviews, our analysis includes a wider range of study designs and intervention strategies, captures more recent and diverse trials, including from under-represented settings, and provides a more nuanced understanding of patient-centred benefits and contextual influences. While prior reviews have established that some models can improve aspects of care,^{15 16} our findings highlight emerging approaches (eg, digital health, transitional care, real-world delivery) and draw attention to subgroups most likely to benefit, such as those with poor baseline control or high socioeconomic disadvantage. Building on these findings, we outline several recommendations to guide policymakers and practitioners in implementing more effective multimorbidity interventions. First, community-based interventions are most effective when targeted to individuals at higher risk, including those with poor baseline control, complex needs or in socioeconomically deprived settings. Second, rather than implementing standalone programmes, policymakers should aim to integrate multimorbidity care into existing primary care structures, supporting coordination through multidisciplinary teams, shared decision-making and case management. Third, digital and telehealth approaches, when tailored to context and supported by adequate infrastructure, can extend the reach of care and improve medication adherence and monitoring. Fourth, routine inclusion of self-management support and patient-centred care models should be prioritised, particularly when designed for cultural and system adaptability. Finally, real-world implementation should account for contextual variability, allowing for local adaptation while maintaining core intervention components. These strategies can help improve outcomes and reduce system inefficiencies associated with fragmented, single-disease care models.

The findings of this review are broadly consistent with previous evidence syntheses, which have similarly reported that community-based interventions for multimorbidity tend to yield greater benefits in patient-reported outcomes than in biomedical markers of disease control.1516 18 19 47,50 Earlier reviews have noted that integrated and collaborative care models are effective for improving mental–physical multimorbidity, particularly depression in the context of diabetes or cardiovascular disease, but their effects on blood pressure, HbA1c and lipid levels are variable and often modest.51,58 One explanation is that patient-centred outcomes, such as quality of life, self-efficacy and satisfaction, are more responsive to interventions that prioritise empowerment, coordination and support, whereas measurable clinical improvements may require longer follow-up, higher intervention fidelity or more intensive tailoring to baseline risk. The subgroup effects observed in our review, such as greater benefit among patients with poor baseline control, those in socioeconomically deprived areas and rural populations, underscore the importance of targeting interventions to those most likely to benefit. From a policy perspective, these findings highlight the potential for community-based models to reduce inequalities in care and outcomes, provided that interventions are adequately resourced, context-sensitive and embedded within primary care systems. For clinical practice, the evidence suggests that while broad implementation of multimorbidity programmes may yield mixed average effects, careful targeting and adaptation can achieve meaningful improvements for high-need populations.

This review has several important strengths. We conducted a comprehensive search across major international and Chinese databases, supplemented by trial registries and grey literature, without time restrictions, and applied rigorous dual-reviewer screening and data extraction in accordance with PRISMA 2020 standards. By including both RCTs and real-world comparative studies, we captured a broad evidence base reflecting both efficacy and effectiveness. The focus on both clinical and patient-centred outcomes provides a more holistic understanding of intervention impact, aligning with the complexity of multimorbidity management. Nevertheless, important limitations should be acknowledged. The quality of the included studies was variable, with frequent risks of bias related to the open-label nature of complex interventions,^{22 24 26 31 32 37 41} potential selection and recruitment bias,^{23 24 37} reliance on self-reported outcomes²² and risks of contamination when providers delivered both intervention and control care.³⁷ Many trials failed to reach their intended sample size^{25 35 37 42} or had substantial attrition,²² resulting in limited statistical power and wide CIs, making it difficult to draw firm conclusions about null effects.

The included studies also showed heterogeneity in their populations, interventions and baseline characteristics. For example, the studies focused on patients with multimorbidity, which is inherently a heterogeneous condition with diverse combinations and severe diseases. Participants varied widely in age, number of chronic conditions and disease burden. One study highlighted that the complexity of multimorbidity makes it challenging to evaluate interventions.⁴¹ Despite randomisation, some studies reported significant baseline differences between intervention and control groups.^{24 25 27 33 35} For example, in one trial, the intervention group had significantly higher blood pressure at baseline.²⁷ In another, the control group had a higher rate of hospital or emergency room admissions in the year prior to the study.⁴⁴ Such imbalances were typically adjusted for in statistical analyses, but they represent a potential source of confounding.^{27 33} Additionally, methodological constraints of this review include the restriction to English and Chinese language publications and the possibility of publication bias despite searching grey literature and registries. Finally, while we applied predefined criteria for study inclusion and extraction, the interpretation of complex interventions in heterogeneous populations inevitably involved some degree of judgement. These limitations suggest that our findings should be interpreted with caution and viewed as indicative of promising directions rather than definitive evidence of effectiveness.

Despite growing interest in interventions for multimorbidity, substantial research gaps remain. The existing literature is marked by inconsistencies, with many trials reporting improvements in patient-centred outcomes^{22 34 41} but neutral or inconclusive findings for clinical markers such as blood pressure, HbA1c and lipid control.^{33 35 37} This heterogeneity reflects both the complexity of multimorbidity and methodological limitations of existing studies. Several populations remain understudied, including younger adults with early multimorbidity, individuals in low-income and middle-income countries where the burden is rapidly increasing, and socioeconomically disadvantaged groups who experience earlier onset and greater severity of multimorbidity. Certain interventions, such as transitional care, medication optimisation⁴⁶ and structured exercise programmes,⁴¹ have been evaluated only in small numbers of studies, while outcomes such as mortality, long-term cardiovascular events and cost-effectiveness remain insufficiently reported.^{29 31} Future research would benefit from adequately powered RCTs and pragmatic real-world evaluations that use standardised definitions of multimorbidity, report both clinical and patient-centred outcomes and include longer follow-up to capture sustained effects. Methodological improvements should also include better recruitment and retention strategies, attention to intervention fidelity and systematic reporting of subgroup analyses to identify who benefits most. Emerging approaches, such as leveraging big data to capture multimorbidity trajectories, artificial intelligence to personalise self-management support and biomarkers to stratify risk, offer promising avenues to advance the field.

In conclusion, this systematic review found that community-based interventions for multimorbidity demonstrate the greatest and most consistent benefits in patient-centred outcomes, including quality of life, self-management and patient experience, with more variable effects on clinical markers such as blood pressure, glycaemic control and lipid levels. Integrated and collaborative care models, digital health interventions and tailored self-management support appear particularly promising, especially for high-risk subgroups with poor baseline control or socioeconomic disadvantages. However, heterogeneities in study design, intervention fidelity and population characteristics limit definitive conclusions. Our findings underscore the need to prioritise patient-centred and context-sensitive models of care that can be embedded in primary and community settings.

Data availability statement

Data sharing not applicable as no datasets generated and/or analysed for this study.