Supporting referral to pulmonary rehabilitation from primary care: A systematic evaluation of interventions using the ERIC taxonomy Authors: Sarah Hug, Sanduni Madawala, Arwel W. Jones, Caitlin Dumergue, Anne E. Holland, Narelle S. Cox Published: 2026 ABSTRACT BackgroundPulmonary rehabilitation (PR) is a key component in managing chronic respiratory diseases (CRDs). Primary care, often the first point of contact for people with CRD, is well-positioned to facilitate referral to PR, yet referral rates are low. This systematic review aimed to identify the key components of primary care interventions to support referral to PR and synthesise their effect on referral rate.MethodsFive electronic databases were searched to identify studies of any design that reported interventions implemented in primary care to support referral to PR for people with CRD. Interventions could target people with CRD and/or healthcare professionals. Screening, quality appraisal using the Downs and Black checklist, and data extraction were conducted independently by two reviewers. Interventions were mapped to the Expert Recommendations for Implementing Change (ERIC) taxonomy—a framework of 73 strategies seeking to enhance the adoption, implementation, and sustainability of evidence-based interventions.ResultsThirteen studies were included, employing a range of quasi-experimental, observational, and randomised designs. Overall, the studies were of moderate quality (mean total score 14, range 10 to 25 out of 27). Interventions incorporated a mean of 12 ERIC strategies across five domains, most commonly education and training, interactive assistance, clinician support, and audit and feedback. Impact of interventions on referral rates was minimal (n = 3 randomised controlled trials, pooled mean difference 0%, 95% confidence interval −0.24 to 0.25).InterpretationMulti-component interventions, including numerous implementation strategies, have achieved minimal improvement in PR referral rates from primary care. Understanding how strategies are delivered and applied, rather than simply the number of strategies or the combination they are used in, may be important for operationalisation of referral focussed interventions. Addressing the behavioural processes underpinning referral decisions, together with the use of theory-informed, context-specific approaches may enhance effectiveness of referral focussed interventions. Introduction Chronic respiratory diseases (CRD) are highly prevalent non-communicable conditions affecting the airways and lung parenchyma1–3. The global burden of CRDs, particularly chronic obstructive pulmonary disease (COPD), is substantial, with global experts underscoring its growing impact and calling for urgent research to address the escalating public health challenge4. While CRDs among adults are incurable, both pharmacological and non-pharmacological treatments can significantly improve quality of life, reduce symptoms, help with overall disease management and mitigate healthcare burden1. Pulmonary rehabilitation (PR), a structured and individualised program combining exercise and education5–7, is the primary non-pharmacological treatment associated with these benefits8–12. Despite being recommended as an important component in the management of CRDs5,13, PR is markedly underutilised14,15, with persistently low referral rates limiting patients’ ability to access and benefit from these programs. The majority of people living with CRD are managed exclusively in primary care16. However rates of referral to PR from primary care are low, often less than 10%15,17,18. Given the evidence for its benefit, low rates of referral are problematic. Healthcare professionals (HCPs) working in primary care identify significant barriers to referral including lack of awareness of the program and referral pathways, low self-efficacy with making a referral and presumed low patient motivation19–23. Additionally, people with CRD report psychological barriers such as fear of breathlessness, uncertainty regarding what the program entails, shame and stigma with accessing healthcare, and transport difficulties hindering the ability to engage or attend19. Previous efforts to increase referrals to PR from primary and tertiary care settings have centred on changing HCPs referral practices—through education, training, computer-generated prompts, and financial incentives—as well as supporting shared decision-making, for example by using tools that encourage discussion during routine reviews24,25, or showing a co-designed video to support patient understanding of the program26. However, effects were at best modest at improving referrals to PR (range 6% to 25%)24. Commonly the interventions under review were part of a larger, multi-component intervention to optimise CRD management, and the effects of individual components on implementation practices were not identified. Identifying which components are most effective for improving referral to PR is important because HCP referral practices and patient acceptance of referral are shaped by multiple interacting determinants—such as knowledge, skills, beliefs about consequences, environmental constraints, and social influences22. To support clearer specification and comparison of intervention components across studies, the Expert Recommendations for Implementing Change (ERIC) taxonomy27 provides a comprehensive list of 73 intervention components, organised into nine domains28, and has demonstrated robust face validity for describing implementation strategies29. This taxonomy enables researchers to identify the active components within each intervention and provides a shared language for examining how implementation strategies are used to support changes in practice. Therefore, the research questions for this systematic review were:i.What are the key components of interventions delivered in primary care to support referral to PR for people with CRD?ii.What is the effect on referral rates of intervention(s) delivered in primary care to support referral to PR? i.What are the key components of interventions delivered in primary care to support referral to PR for people with CRD? What are the key components of interventions delivered in primary care to support referral to PR for people with CRD? ii.What is the effect on referral rates of intervention(s) delivered in primary care to support referral to PR? What is the effect on referral rates of intervention(s) delivered in primary care to support referral to PR? Methods This systematic review was prospectively registered (PROSPERO CRD42024595844) and is reported according to the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) guidelines30. Identification and selection of studies We searched Medline, Cinahl, PEDro, EMBASE (via Ovid) and PsycINFO from inception to 13 November 2025. The search strategy was developed in consultation with a librarian (see Supplement 1). We sought to find primary research articles of any trial design (excluding case studies), published as full-text articles in peer-reviewed journals that report on an intervention implemented in primary care to support referral to PR for adults with CRD. The interventions could target people with CRDs, HCPs who refer to PR, or both. Studies published in languages other than English were excluded. Studies of interventions that included the delivery of PR in a primary care setting, rather than referral to PR, were also excluded. Study screening Search results were uploaded to Covidence to support study screening. Duplicates were removed and two reviewers (SH, SM, NSC, AWJ or CD) independently screened title and abstracts for eligibility. Studies that could not be excluded based on reviewing title and abstracts proceeded to independent screening of the full text, by two reviewers, to determine final eligibility. Any conflicts in full text screening were managed by discussion and involvement of a third review author as required. Data extraction and quality assessment Data pertaining to study characteristics, the intervention, mode of delivery, target group, and referral rates were extracted from each study. Using a data extraction template of the ERIC taxonomy24, two reviewers independently coded ERIC components applied in each intervention, which was checked and collated by a third reviewer. Methodological quality was assessed independently by two reviewers using the Downs and Black checklist25. The checklist allows for comparison of quality across different study designs, and assesses study quality, external validity, study bias, confounding and selection bias and power of the study. An overall quality score, and scores for each of the five sections was determined. Quality scoring discrepancies were resolved by involving a third reviewer. Data management and analysis The ERIC taxonomy was used to identify and map the implementation components applied in each intervention. To understand the implementation target of each intervention, ERIC domains were grouped according to macro, meso- or micro-levels. Macro-level grouping indicated domains that act at the health-system level through policy, funding or infrastructure; meso-level grouping indicated domains that support how services and professionals work together to deliver care; and micro-level grouping indicated domains that act at the individual HCP–patient interface to shape referral decisions. ERIC domain and component use were reported narratively. To examine how ERIC components were used together across studies, we calculated proportional co-occurrence between individual ERIC implementation components. For each component, we identified all studies in which it appeared and then calculated the proportion of those studies in which a second component also appeared. This normalisation accounts for large differences in how frequently individual components were used and provides a comparable measure of association strength (i.e., “given component X was used, how often was component Y also used?”). Co-occurrence analyses and visualisations were conducted in R (version 4.5.1)26. Referral rates associated with the interventions were extracted as reported by study authors. Where possible, we planned to tabulate all referral data as total number of referrals (count data) or number/proportion of participants referred (dichotomous data). Where sufficient data were available from randomised controlled trials, results were meta-analysed in RevMan 527, and a pooled mean difference with 95% confidence intervals calculated. Results The flow of studies through this review is summarised in Fig. 1. Searches detected 1908 records. After removal of duplicates, 1426 title and abstracts were screened. 66 records progressed to full text review, of which 13 studies were included in this review28,29,31–41. Included studies were conducted in Australia, the United Kingdom, Denmark, and Sweden, comprising a mix of quasi-experimental pre–post designs, observational cohorts, mixed-methods approaches, and cluster randomised controlled trials. Of the included studies, 12 had interventions that were designed to change referral practices of primary care HCPs (e.g., general practitioners, practice nurses, pharmacists or physiotherapists). Only one study targeted the person with CRD with an intervention that encouraged shared-decision making with the general practitioner (GP)38. Table 1 summarises the characteristics of included studies.Fig. 1Flow of studies through the review30.Table 1Characteristics of included studies.Study referenceStudy designSettingIntervention target groupIntervention and deliveryParticipantsAngus et al.28Observational, feasibility studyUnited KingdomPractice nurses (n = 18)NICE Guideline-based computer-guided consultation designed to facilitate COPD managementDelivery timeframe not stated191 people with COPD with full computer-guided consultsDeprez et al.29Quasi-experimental (pre-post)United States of AmericaPrimary care physicians and mid-level providersb (n = 25)Breakthrough series model educational sessions and clinical targets.Delivery timeframe not statedData from 584 COPD patient charts were collected in the pre-initiative phase and from 626 patient charts in the post-initiative phaseFathima et al.31Quasi-experimental (pre-post)AustraliaConsultant pharmacists (n = 2)Pharmacist-led COPD screening and management through community pharmacies. The service consisted of inhaler technique assessment; medication adherence assessment; and referrals to the patient’s GP to facilitate the uptake of non-pharmacological resources including PR.3 in-pharmacy visits and 2 follow-up phone calls over 6-month period40 people with COPD recruited from nine host pharmaciesFoster et al.32Mixed methods participatory action researchUnited KingdomGPs (n = 9) and practice nurses (n = 13) from eight participating practices- In house PR training and education sessions run by PR representatives focussing on discussions about PR, referral processes and the local services available- Integration of clear prompts for referral on COPD review template- Creation of protocols for referral- Patient information materials- Memory aids (mugs and coasters)- Audit and feedback of referral data- Surveying people with COPD about their referral experienceDelivery timeframe not statedSix practices provided patient data for the audit. 126 participants with COPD filled in the questionnaire about their referral experience. Number of participants audited for PR referral not givenHull et al33Longitudinal auditLondon36 general practicesa- COPD integrated care packaged based on the NICE COPD guidelines.- Community respiratory team support- COPD review template and care- Financial incentives to refer- Quarterly MDT case reviews- Monthly update electronic dashboard of key performance measures (including referral to PR)Evaluated across 3 yearsAll 36 practices provided patient data from their COPD registersLange et al.34Quasi-experimental (pre–post)DenmarkGeneral practitioners (n = 154)National GP educational program including:- workshops on guidelines- forum for case discussion- audit of COPD care records 1 year before and after education program2549 patient records included in first survey, 2394 in the secondMartinez et al. 202435Cluster RCTUnited States of AmericaPrimary care physicians (n = 100)- Basic COPD education (30-minute video based on GOLD guidelines)- CAPTURE screening tool training (use and interpretation)- Access to patient screening dataDelivery timeframe not stated2454 participants with COPD in intervention group2210 in control groupPagano et al.36Pilot observationalAustralia4 general practicesaPhysiotherapistCardiorespiratory physiotherapist deployed into general practice for partnership with GP to:- refer to PR- provide physical activity advice and a booklet- provide a pedometer- deliver smoking cessation advice- review inhaler technique- provide COPD action planDelivered across a 3-month window40 participants with COPD completed the intervention with the physiotherapist. Due to attrition, 27 patient records audited for PR referralRoberts et al.37Prospective observational cohort study with service evaluationUnited KingdomGP’s and practice nurses- Integrated COPD management strategy (accurate diagnosis, smoking cessation, PR referral)- Educational programme comprising monthly workshops, study days, online education, increased access to consultant-led MDT clinics- clinical auditEvaluated at 12-months following intervention5491 people with COPD (from 55 practices)Roberts et al.38Quasi-experimental (non-randomised control trial)United KingdomPeople with COPD and GPsPatient-held co-designed COPD quality care scorecard to be brought in and discussed with GPs at annual COPD reviews- The scorecard mapped indicators from NICE quality standards to data from patients’ clinical records. Developed collaboratively with patients to empower them in discussions with HCPs.- The control group received standard COPD management without any scorecard or additional feedback mechanismOutcomes compared before and 3 months after the scorecard was implemented1235 people with COPD (640 intervention and 595 control)Sandelowsky et al.39Cluster RCTSwedenGeneral practitioners (n = 24)Case method learning versus traditional lecture learningDelivery timeframe not stated425 people with COPD (case-method group n = 209, traditional-lecture group n = 216) known to one of the 24 general practicesUlrik et al.40Quasi-experimental (pre-post)DenmarkParticipating GPs (n = 124) and their staff, including nurses, laboratory technicians, and secretariesEducational program on COPD managementDelivery timeframe not stated1716 patients with COPD included in the pre-survey, 1342 in the post-surveyZwar et al.41Cluster RCTAustraliaGeneral practitioners and practice nursesPractice nurse and GP COPD management including:- training- electronic care plan template- PR referrals- inhaler technique, smoking cessation and self-management adviceDelivery timeframe not stated126 participants with COPD randomised to the intervention group, 96 to the control groupCAPTURE, COPD Assessment in Primary Care To Identify Undiagnosed Respiratory Disease and Exacerbation Risk; COPD, chronic obstructive pulmonary disease, GOLD Global obstructive lung disease strategy, GP general practitioner, NICE national institute for health and medical excellence, MDT, multidisciplinary team, PR pulmonary rehabilitation, RCT randomised controlled trial.anumber of GP’s involved in this intervention was unclear.bMid-level practitioners are non-physician healthcare providers who assess, diagnose, and treat patients, but do not have the full training or licensure of physicians. Examples include nurse practitioners and physician assistants who work alongside primary care physicians. Flow of studies through the review30. Table 1Characteristics of included studies.Study referenceStudy designSettingIntervention target groupIntervention and deliveryParticipantsAngus et al.28Observational, feasibility studyUnited KingdomPractice nurses (n = 18)NICE Guideline-based computer-guided consultation designed to facilitate COPD managementDelivery timeframe not stated191 people with COPD with full computer-guided consultsDeprez et al.29Quasi-experimental (pre-post)United States of AmericaPrimary care physicians and mid-level providersb (n = 25)Breakthrough series model educational sessions and clinical targets.Delivery timeframe not statedData from 584 COPD patient charts were collected in the pre-initiative phase and from 626 patient charts in the post-initiative phaseFathima et al.31Quasi-experimental (pre-post)AustraliaConsultant pharmacists (n = 2)Pharmacist-led COPD screening and management through community pharmacies. The service consisted of inhaler technique assessment; medication adherence assessment; and referrals to the patient’s GP to facilitate the uptake of non-pharmacological resources including PR.3 in-pharmacy visits and 2 follow-up phone calls over 6-month period40 people with COPD recruited from nine host pharmaciesFoster et al.32Mixed methods participatory action researchUnited KingdomGPs (n = 9) and practice nurses (n = 13) from eight participating practices- In house PR training and education sessions run by PR representatives focussing on discussions about PR, referral processes and the local services available- Integration of clear prompts for referral on COPD review template- Creation of protocols for referral- Patient information materials- Memory aids (mugs and coasters)- Audit and feedback of referral data- Surveying people with COPD about their referral experienceDelivery timeframe not statedSix practices provided patient data for the audit. 126 participants with COPD filled in the questionnaire about their referral experience. Number of participants audited for PR referral not givenHull et al33Longitudinal auditLondon36 general practicesa- COPD integrated care packaged based on the NICE COPD guidelines.- Community respiratory team support- COPD review template and care- Financial incentives to refer- Quarterly MDT case reviews- Monthly update electronic dashboard of key performance measures (including referral to PR)Evaluated across 3 yearsAll 36 practices provided patient data from their COPD registersLange et al.34Quasi-experimental (pre–post)DenmarkGeneral practitioners (n = 154)National GP educational program including:- workshops on guidelines- forum for case discussion- audit of COPD care records 1 year before and after education program2549 patient records included in first survey, 2394 in the secondMartinez et al. 202435Cluster RCTUnited States of AmericaPrimary care physicians (n = 100)- Basic COPD education (30-minute video based on GOLD guidelines)- CAPTURE screening tool training (use and interpretation)- Access to patient screening dataDelivery timeframe not stated2454 participants with COPD in intervention group2210 in control groupPagano et al.36Pilot observationalAustralia4 general practicesaPhysiotherapistCardiorespiratory physiotherapist deployed into general practice for partnership with GP to:- refer to PR- provide physical activity advice and a booklet- provide a pedometer- deliver smoking cessation advice- review inhaler technique- provide COPD action planDelivered across a 3-month window40 participants with COPD completed the intervention with the physiotherapist. Due to attrition, 27 patient records audited for PR referralRoberts et al.37Prospective observational cohort study with service evaluationUnited KingdomGP’s and practice nurses- Integrated COPD management strategy (accurate diagnosis, smoking cessation, PR referral)- Educational programme comprising monthly workshops, study days, online education, increased access to consultant-led MDT clinics- clinical auditEvaluated at 12-months following intervention5491 people with COPD (from 55 practices)Roberts et al.38Quasi-experimental (non-randomised control trial)United KingdomPeople with COPD and GPsPatient-held co-designed COPD quality care scorecard to be brought in and discussed with GPs at annual COPD reviews- The scorecard mapped indicators from NICE quality standards to data from patients’ clinical records. Developed collaboratively with patients to empower them in discussions with HCPs.- The control group received standard COPD management without any scorecard or additional feedback mechanismOutcomes compared before and 3 months after the scorecard was implemented1235 people with COPD (640 intervention and 595 control)Sandelowsky et al.39Cluster RCTSwedenGeneral practitioners (n = 24)Case method learning versus traditional lecture learningDelivery timeframe not stated425 people with COPD (case-method group n = 209, traditional-lecture group n = 216) known to one of the 24 general practicesUlrik et al.40Quasi-experimental (pre-post)DenmarkParticipating GPs (n = 124) and their staff, including nurses, laboratory technicians, and secretariesEducational program on COPD managementDelivery timeframe not stated1716 patients with COPD included in the pre-survey, 1342 in the post-surveyZwar et al.41Cluster RCTAustraliaGeneral practitioners and practice nursesPractice nurse and GP COPD management including:- training- electronic care plan template- PR referrals- inhaler technique, smoking cessation and self-management adviceDelivery timeframe not stated126 participants with COPD randomised to the intervention group, 96 to the control groupCAPTURE, COPD Assessment in Primary Care To Identify Undiagnosed Respiratory Disease and Exacerbation Risk; COPD, chronic obstructive pulmonary disease, GOLD Global obstructive lung disease strategy, GP general practitioner, NICE national institute for health and medical excellence, MDT, multidisciplinary team, PR pulmonary rehabilitation, RCT randomised controlled trial.anumber of GP’s involved in this intervention was unclear.bMid-level practitioners are non-physician healthcare providers who assess, diagnose, and treat patients, but do not have the full training or licensure of physicians. Examples include nurse practitioners and physician assistants who work alongside primary care physicians. Characteristics of included studies. NICE Guideline-based computer-guided consultation designed to facilitate COPD management Delivery timeframe not stated Breakthrough series model educational sessions and clinical targets. Delivery timeframe not stated Pharmacist-led COPD screening and management through community pharmacies. The service consisted of inhaler technique assessment; medication adherence assessment; and referrals to the patient’s GP to facilitate the uptake of non-pharmacological resources including PR. 3 in-pharmacy visits and 2 follow-up phone calls over 6-month period - In house PR training and education sessions run by PR representatives focussing on discussions about PR, referral processes and the local services available - Integration of clear prompts for referral on COPD review template - Creation of protocols for referral - Patient information materials - Memory aids (mugs and coasters) - Audit and feedback of referral data - Surveying people with COPD about their referral experience Delivery timeframe not stated - COPD integrated care packaged based on the NICE COPD guidelines. - Community respiratory team support - COPD review template and care - Financial incentives to refer - Quarterly MDT case reviews - Monthly update electronic dashboard of key performance measures (including referral to PR) Evaluated across 3 years National GP educational program including: - workshops on guidelines - forum for case discussion - audit of COPD care records 1 year before and after education program - Basic COPD education (30-minute video based on GOLD guidelines) - CAPTURE screening tool training (use and interpretation) - Access to patient screening data Delivery timeframe not stated 2454 participants with COPD in intervention group 2210 in control group 4 general practicesa Physiotherapist Cardiorespiratory physiotherapist deployed into general practice for partnership with GP to: - refer to PR - provide physical activity advice and a booklet - provide a pedometer - deliver smoking cessation advice - review inhaler technique - provide COPD action plan Delivered across a 3-month window - Integrated COPD management strategy (accurate diagnosis, smoking cessation, PR referral) - Educational programme comprising monthly workshops, study days, online education, increased access to consultant-led MDT clinics - clinical audit Evaluated at 12-months following intervention Patient-held co-designed COPD quality care scorecard to be brought in and discussed with GPs at annual COPD reviews - The scorecard mapped indicators from NICE quality standards to data from patients’ clinical records. Developed collaboratively with patients to empower them in discussions with HCPs. - The control group received standard COPD management without any scorecard or additional feedback mechanism Outcomes compared before and 3 months after the scorecard was implemented Case method learning versus traditional lecture learning Delivery timeframe not stated Educational program on COPD management Delivery timeframe not stated Practice nurse and GP COPD management including: - training - electronic care plan template - PR referrals - inhaler technique, smoking cessation and self-management advice Delivery timeframe not stated CAPTURE, COPD Assessment in Primary Care To Identify Undiagnosed Respiratory Disease and Exacerbation Risk; COPD, chronic obstructive pulmonary disease, GOLD Global obstructive lung disease strategy, GP general practitioner, NICE national institute for health and medical excellence, MDT, multidisciplinary team, PR pulmonary rehabilitation, RCT randomised controlled trial. anumber of GP’s involved in this intervention was unclear. bMid-level practitioners are non-physician healthcare providers who assess, diagnose, and treat patients, but do not have the full training or licensure of physicians. Examples include nurse practitioners and physician assistants who work alongside primary care physicians. Quality assessment A summary of quality assessment is provided in Table 2. All studies had clearly described aims, outcomes and results with reasonable external validity. Outcome assessment was often not blinded, and only three studies were of a randomised controlled trial design33,37,39, therefore randomisation and concealment were not applicable to the remaining studies. Full details of the quality assessment are provided in the supplement (Table S1).Table 2Quality assessment using the Downs and Black checklist25.SectionAngus 2012Deprez 2009Fathima 2021Foster 2016Hull 2024Lange 2007Martinez 2024Pagano 2023Roberts 2010Roberts 2015Sandelowsky 2020Ulrik 2010Zwar 2016Reporting quality7696479977979External validity2211233222111Study bias2152236422335Internal validity (confounding/ selection)1112326211213Power0010001000101Overall12/27a10/27a17/2711/27a11/27a15/27a25/27a17/27a12/27a12/27a16/2712/27a19/27aatotal score had items that were not applicable. Table 2Quality assessment using the Downs and Black checklist25.SectionAngus 2012Deprez 2009Fathima 2021Foster 2016Hull 2024Lange 2007Martinez 2024Pagano 2023Roberts 2010Roberts 2015Sandelowsky 2020Ulrik 2010Zwar 2016Reporting quality7696479977979External validity2211233222111Study bias2152236422335Internal validity (confounding/ selection)1112326211213Power0010001000101Overall12/27a10/27a17/2711/27a11/27a15/27a25/27a17/27a12/27a12/27a16/2712/27a19/27aatotal score had items that were not applicable. Quality assessment using the Downs and Black checklist25. atotal score had items that were not applicable. Intervention components according to ERIC taxonomy Of the included studies, the average number of ERIC intervention components and domains used in the intervention of each study was 12 (range 4 to 25) and five (range 2 to 8), respectively (Fig. 2, Table S2). Figure 3 gives examples of interventions coded to ERIC domains within the macro-, meso- and micro groupings. The most frequently used intervention components came from the ‘train and educate stakeholders’ and ‘provide interactive assistance’ domains and included: conducting educational meetings (n = 11), ongoing training (n = 6), making training dynamic (n = 5), providing ongoing consultations (n = 5), providing local technical assistance (n = 6) and changing record systems (n = 6 [Table S2]). Patterns of how ERIC implementation components were combined within interventions were explored using proportional co-occurrence heatmaps (Figures S1 to S9). These visualisations show, for each ERIC domain (columns), the extent to which its individual components tended to be used alongside individual components from other domains (rows), independent of how frequently each component appeared overall. Stronger associations (darker shading) indicate components that were consistently implemented together when they were used. For example, components from the ‘train and educate stakeholders’ (Figure S1) and ‘provide interactive assistance’ (Figure S2) domains frequently co-occurred with ‘use evaluative and iterative’ components such as audit and feedback. Whereas components from the ‘financial strategies’ (Figure S6) domain showed minimal co-occurrence with other strategies, reflecting their limited use across studies.Fig. 2ERIC taxonomy domain used across to support referral to PR.Fig. 3Example of intervention components of individual studies coded to ERIC domains.Footnotes: Macro-level grouping indicates domains that act at the health-system level through policy/funding/infrastructure; meso-level grouping indicates domains that support how services and professionals work together to deliver care; and micro-level grouping indicates domains that act at the individual HCP–patient interface to shape referral decisions. ERIC taxonomy domain used across to support referral to PR. Example of intervention components of individual studies coded to ERIC domains. Footnotes: Macro-level grouping indicates domains that act at the health-system level through policy/funding/infrastructure; meso-level grouping indicates domains that support how services and professionals work together to deliver care; and micro-level grouping indicates domains that act at the individual HCP–patient interface to shape referral decisions. Intervention effect on referral to PR Data from three randomised controlled trials35,39,41 were pooled in a meta-analysis showing that, compared with the control group, the impact of interventions on referral rates was minimal (pooled mean difference 0%, 95% confidence interval −0.24% to 0.25%, Figure S10). Seven non-randomised studies provided pre- and post-intervention referral data29,31,33,34,37,38,40, showing an increase from an average referral rate of 15.3% pre-intervention (range: 1.4% to 45.6%, Table 3), to 21.3% post-intervention (range: 0.8% to 60.1%, Table 3). Three studies28,32,36 reported post-intervention data only, but did not provide the total number of participants eligible for referral. Among these studies, referral proportions ranged between 24% to 100% (Table 3). Of the studies that reported improvements in referral rate after the intervention was implemented (increase in referral rate ranging 4 to 25%), the greatest gains occurred in those that applied multiple ERIC components in combination such as education together with interaction, support and feedback mechanisms, across several domains29,33,34,38.Table 3Summary of change in referral to pulmonary rehabilitation pre- to post-implementation of the intervention in n = 12 included studies.StudyStudy designReferral rate pre-implementationReferral rate post-implementationBetween group difference (95%CI) / P valueAngus et al.28Observational, feasibility studyNot reported47 (24%) were referred to pulmonary rehabilitationNot applicableDeprez et al.29Quasi-experimental (pre-post)7%12%P = 0.048Fathima et al.31Quasi-experimental (pre-post)n = 10 (27%)n = 10 (37%)P = 0.39Foster et al.32Mixed methods participatory action researchNot reportedRange 27% to 100%Not applicableHull et al.33Longitudinal audit45.6%Year 1 = 60.1%Year 2 = 70.7%Not applicableLange et al.34Quasi-experimental (pre–post)16.7%20.2%P < 0.01Martinez et al. 202435Cluster RCTIntervention group referrals to PR = 6.4%, control group referrals to PR = 7% 0.6% (95% CI −6.4 to 5.1)Pagano et al.36Pilot observationalNot reportedn = 21 (78%)Not applicableRoberts et al.37Prospective observational cohort study with service evaluationn = 84n = 143Not applicableRoberts et al.38Quasi-experimental (non-randomised control trial)Intervention group = 10 (1.7%)Control group = 15 (2%)Intervention group = 51 (7.8%)Control group = 37 (4.6%)P = 0.03Sandelowsky et al.39Cluster RCTIntervention group = 10.6%Control group = 15.7%Intervention group = 16.8%Control group = 18.8% 3.0% (95% CI −5.5 to 11.6)Ulrik et al.40Quasi-experimental (pre-post)12%16%P < 0.01Zwar et al.41Cluster RCTIntervention group = 2 (1.4%)Control group = 0Intervention group = 1 (0.8%)Control group = 1 (1%)0% (95% CI −0.24 to 0.24)Data is reported according to how it was presented in the original studies.CI confidence interval. Table 3Summary of change in referral to pulmonary rehabilitation pre- to post-implementation of the intervention in n = 12 included studies.StudyStudy designReferral rate pre-implementationReferral rate post-implementationBetween group difference (95%CI) / P valueAngus et al.28Observational, feasibility studyNot reported47 (24%) were referred to pulmonary rehabilitationNot applicableDeprez et al.29Quasi-experimental (pre-post)7%12%P = 0.048Fathima et al.31Quasi-experimental (pre-post)n = 10 (27%)n = 10 (37%)P = 0.39Foster et al.32Mixed methods participatory action researchNot reportedRange 27% to 100%Not applicableHull et al.33Longitudinal audit45.6%Year 1 = 60.1%Year 2 = 70.7%Not applicableLange et al.34Quasi-experimental (pre–post)16.7%20.2%P < 0.01Martinez et al. 202435Cluster RCTIntervention group referrals to PR = 6.4%, control group referrals to PR = 7% 0.6% (95% CI −6.4 to 5.1)Pagano et al.36Pilot observationalNot reportedn = 21 (78%)Not applicableRoberts et al.37Prospective observational cohort study with service evaluationn = 84n = 143Not applicableRoberts et al.38Quasi-experimental (non-randomised control trial)Intervention group = 10 (1.7%)Control group = 15 (2%)Intervention group = 51 (7.8%)Control group = 37 (4.6%)P = 0.03Sandelowsky et al.39Cluster RCTIntervention group = 10.6%Control group = 15.7%Intervention group = 16.8%Control group = 18.8% 3.0% (95% CI −5.5 to 11.6)Ulrik et al.40Quasi-experimental (pre-post)12%16%P < 0.01Zwar et al.41Cluster RCTIntervention group = 2 (1.4%)Control group = 0Intervention group = 1 (0.8%)Control group = 1 (1%)0% (95% CI −0.24 to 0.24)Data is reported according to how it was presented in the original studies.CI confidence interval. Summary of change in referral to pulmonary rehabilitation pre- to post-implementation of the intervention in n = 12 included studies. Year 1 = 60.1% Year 2 = 70.7% Intervention group = 10 (1.7%) Control group = 15 (2%) Intervention group = 51 (7.8%) Control group = 37 (4.6%) Intervention group = 10.6% Control group = 15.7% Intervention group = 16.8% Control group = 18.8% Intervention group = 2 (1.4%) Control group = 0 Intervention group = 1 (0.8%) Control group = 1 (1%) Data is reported according to how it was presented in the original studies. CI confidence interval. Discussion While previous reviews15 have established that referrals to PR from primary care remain suboptimal and that existing interventions achieve minimal gains23, little attention has been given to examining how interventions are constructed and which implementation components may contribute to observed changes in referral behaviours. In addition to synthesising interventions to optimise referral to PR for people with CRD, this review is the first to apply the ERIC taxonomy to systematically document their implementation components. We identified that ERIC components were rarely used in isolation, and studies that demonstrated the greatest, albeit modest, improvements in referral rates29,33,34,38 combined a variety of ERIC components across domains including education and training for HCPs, interactive assistance, clinician support and data feedback/prompts. Interventions primarily focussed at the meso- and micro-levels, targeting referral practices of HCPs and, overall, may have resulted in small increases in PR referral rates. However, the certainty of effect size was low for RCTs, and risk of bias was high for non-RCT studies. Despite knowing what implementation strategies were applied, studies did not give information on how these interventions were delivered, leaving little understanding of their operationalisation and fidelity. There are also a number of ERIC components that were not applied, and their utility in supporting PR referral is unclear (Figures S1 to S9). As a result, it remains difficult to determine which implementation components were responsible for observed changes in referral practices, or why similar strategies produced variable effects across settings. HCP training and education were the most frequently employed ERIC component across included studies (85%); however, their widespread use did not translate into large or consistent improvements in PR referral rates. This suggests that while HCP education may be a necessary component of interventions to support referral to PR, it is unlikely to be sufficient when used in isolation to change referral. How these educational strategies were operationalised was rarely reported in the included papers. In many studies, educational components focused on disseminating guideline-based knowledge; however, the mode of education delivery and content, as well as the time taken to deliver these strategies (e.g., dose) were not described. Recent qualitative work suggests that referral decisions are shaped by the quality and degree of engagement during a HCP–patient interaction, with successful referrals typically emerging from two-way conversations that provide nuanced information, identify patient goals, and address concerns to support shared decision-making19,42,43. Strategies in included studies did not address how HCPs communicate about PR or navigate the referral discussion with patients. While earlier work has highlighted the importance of HCP–patient communication in supporting referral to PR19,44, how to intentionally support these interactions within interventions to improve referral to PR requires exploration. Other commonly used ERIC components in interventions were audit and feedback and facilitation of clinical data to providers; however, their application was typically limited to reporting referral rates back to providers at the conclusion of the study. While data feedback can raise awareness of referral practices and gaps, its focus on outcomes rather than the process underpinning referral practices may limit potential for change in referral behaviours. Data feedback could alternatively be operationalised by PR clinicians closing the loop to referrers by confirming whether referrals were received, actioned, and providing post-rehabilitation patient summaries. Doing so may increase the capacity of referrers to reinforce and sustain referral practices. Organisational and healthcare system level strategies such as referral prompts and record-system changes were also used. However, these strategies varied in how explicitly they supported referrals, with many focusing on reminding HCPs to refer rather than clarifying what information was required or embedding referral processes seamlessly into routine care. Collectively, these findings suggest that while ERIC components were commonly implemented, their operationalisation may not have consistently targeted the behavioural processes required to support sustained change in referral practices. Future work may benefit from explicitly linking implementation strategies (e.g., ERIC) with behaviour change frameworks45 to guide both strategy selection and their operationalisation within specific clinical contexts46. Across included studies, interventions primarily focussed at the meso- and micro-levels, with a focus on modifying HCP referral practices through education and training, clinician support and clinical audits. While identifying macro-level levers were not the focus of this review, they, as well as patient-facing strategies, were rare. Only one study38 delivered an intervention that explicitly engaged both patients and HCPs through a patient-held, co-designed scorecard that encouraged shared decision-making during consultations. Compared with control practices, those that utilised the scorecard demonstrated an improvement in PR referrals (intervention, 1.7% to 7.8%; control, 2.0% to 4.6%), highlighting the value of directly involving patients in the referral process. Given that patients must actively engage with and commit to PR, interventions that overlook the patient perspective at the time of referral may miss an important opportunity to support PR referral and uptake. These findings support the need for future interventions to look holistically, and partner with both patients with CRD and HCPs, to co-design interventions that address their individual barriers to referral47,48. The importance of macro-level levers for improving referral rates to PR, as have been proposed for cardiac rehabilitation49, may also warrant exploration. This review has several strengths. First, mapping interventions to the ERIC taxonomy enabled a standardised analysis of implementation strategies used to support referral to PR from primary care, providing clarity about what intervention components were delivered and facilitating comparison across studies. This structured approach highlights opportunities for more targeted, theory-informed implementation design. In addition, the inclusion of diverse study designs reflects real-world implementation contexts, enhancing the applicability and generalisability of findings across primary care settings. Some limitations should also be acknowledged. The methodological quality of included studies varied, and heterogeneity in study design and referral rate reporting limited the ability to directly compare effects across interventions. Furthermore, reporting of PR referral outcomes across included studies was inconsistent and interpretation was complicated by variability in how referral was defined, measured, and reported. Several studies26,29,34 relied on single post-intervention observational time points and did not report the number of patients eligible for referral, limiting the ability to quantify change over time and compare effects between settings. In a number of studies, referral to PR was embedded within broader packages of COPD care, making it difficult to attribute observed changes in referral rates specifically to referral-focused intervention components. While this reflects the complexity of real-world clinical practice—where PR referral is rarely delivered in isolation—it underscores the need for future interventions to more clearly specify referral as a behavioural target and to align implementation strategies and outcome measures accordingly. Conclusion Referral to PR remains suboptimal, and interventions to support referral from primary care have generally achieved minimal improvements. By applying the ERIC taxonomy to systematically map intervention components, this review indicates that the effectiveness of referral-focused interventions may depend not only on the selection of implementation strategies, but also the combinations they are used in. Irrespective of what combination of implementation components were used, details about how strategies were operationalised to shape referral decisions in primary care need to be better articulated. Future interventions to improve referral to PR should consider involvement of all stakeholders in the referral process. Moving beyond strategy selection alone and adopting theory-informed, context-specific approaches that deliberately link implementation strategies with locally relevant behavioural determinants may support meaningful and sustained improvements in PR referral practices from primary care. Supplementary information Supplementary material Fig_s1 Fig_s2 Fig_s3 Fig_s4 Fig_s5 Fig_s6 Fig_s7 Fig_s8 Fig_s9 Fig_s10 Supplementary material Fig_s1 Fig_s2 Fig_s3 Fig_s4 Fig_s5 Fig_s6 Fig_s7 Fig_s8 Fig_s9 Fig_s10