Effectiveness of Interventions to Reduce Carbon-Emissions Within Secondary Healthcare: Systematic Review and Evidence and Gap Map

Background

The National Health Service (NHS) in England wants to reduce the amount of carbon they produce because it currently contributes 4% to England's overall carbon emissions.

What We Want to Know?

The UK government's Greener NHS team from NHS England asked the Exeter Policy Research Programme Evidence Review Facility to bring together and analyse research which assesses different ways carbon emissions resulting from hospital-led care can be reduced, without affecting the care patients receive in hospitals, at home and in clinics.

Description of the Problem

‘The climate emergency is also a health emergency’ (England 2024). Climate change directly impacts the health of the human population through events such as earthquakes, flooding, heatwaves and drought, which increase the risk of injury, displacement, disruption of food supplies, infectious diseases and mental ill health (England 2024; Lenzen et al. 2020; Tennison et al. 2021; The Lancet Respiratory Medicine 2023). The impact on population health of these climate events, alongside indirect health consequences such as increased prevalence of respiratory conditions due to air pollution, places an increased burden on health services (Royal College of Physicians 2017). This health burden disproportionately impacts populations of lower-income countries, despite contributing a lower proportion to global greenhouse gas (GHG) emissions than developed countries (Bhargawa Ruma 2023; Levy and Patz 2015; Schöngart et al. 2025), in part due to increased exposure risk of exposure to climate events, fewer resources available to support adaptation (Schöngart et al. 2025), and exacerbating existing barriers to accessing health services (Martins et al. 2024; Naser et al. 2024). Concerns have been raised regarding whether high-income countries are doing enough to meet decarbonisation obligations agreed to as part of the Paris Agreement 2015 (United Nations (Paris Agreement) 2015; Vogel and Hickel 2023; Barratt 2022; Brownlee et al. 2017; Lenzen et al. 2020; Or and Seppänen 2024).

Why It Is Important to Do This Review

While necessary for improving and maintaining human well-being, the environmental footprint of healthcare services contributes between 1% and 5% towards total global environmental impacts (Lenzen et al. 2020; Tennison et al. 2021). If the healthcare system were a country, it would be the fifth largest producer of GHGs in the world (Barratt 2022), with contributions towards climate change expected to increase as demand for care increases worldwide (Or and Seppänen 2024).

Reducing the impact of the healthcare system on climate change has the potential to benefit population health through improved air quality and diet and increased activity levels (Mailloux et al. 2021). Within UK healthcare, common sources of carbon emissions between 1990 and 2019 included supply chains (62%), and other sources including delivery of care (24%), travel to and from sites by staff, patients and visitors (10%) and private health and care services commissioned by the NHS (4%) (Tennison et al. 2021). In addition to these, other factors contributing to the environmental footprint of global health services include the overuse of medical services, such as unnecessary tests or treatments (Brownlee et al. 2017), and heat and electricity use (Lenzen et al. 2020). These environmental hotspots provide opportunities for targeted interventions to reduce the climate impact of health services.

Within the United Kingdom, the NHS has exceeded its commitments under the Climate Change Act by reducing its carbon footprint by 30% (Climate Change Act 2008; England 2024). The NHS England Greener NHS, alongside the Primary Care and Medicines policy teams, have been working closely with patients, clinicians and industry to minimise emissions from medicines and anaesthetic gases, reducing waste, ensuring the right medicines are available to patients, and finding mechanisms to support shared, informed decision making (NHS England 2022). Other work includes focusing on identifying and delivering interventions to reduce carbon emissions within known carbon hotspots, such as estates and facilities, travel and transport, supply chain, and certain medicines and medical and anaesthetic gases that have high global warming potential is already underway, alongside examining the effectiveness of different models of care delivery across all specialities, to enable safe, patient-centred lower carbon care models (NHS England 2022; NHS Shared Business Services 2022).

Evidence focusing on the effectiveness of interventions to reduce carbon emissions within secondary healthcare would be a useful complement to work already underway both within the United Kingdom and other high-income countries. An approach which considers how care is delivered across patient care pathways within individual specialities may help support the delivery of equitable and accessible high-quality care. Such an approach can also consider other wider health services policies and ensure that all those who are involved in the design and delivery of care are involved. This review was commissioned by the Greener NHS team and could serve as a useful case study for wider net-zero ambitions elsewhere in the world.

Existing Systematic Review Evidence

Scoping of the evidence base indicates that there are several systematic reviews which examine different types of interventions to reduce carbon emissions, which are summarised below.

Four systematic reviews focus on interventions to reduce carbon emissions within operating theatres (Keil et al. 2023; Papadopoulou et al. 2022; Perry et al. 2023; Siu et al. 2017). Papadopoulou et al. (2022) examine the environmental sustainability of minimally invasive surgery techniques (including robotic and laparoscopic surgery) and include studies from a variety of different specialities which examine different interventions such as cost-awareness campaigns and reusable instruments or report a life cycle assessment (LCA) for a particular surgical procedure (Papadopoulou et al. 2022). The number of studies evaluating/modelling the effect of an intervention in this review was limited (n = 6), with gynaecology and gastroenterology being the main surgical specialities represented (Papadopoulou et al. 2022). In the review conducted by Perry et al. (2023), studies evaluated interventions focusing on recycling and waste management, waste reduction, reuse, reprocessing/LCA, energy and resource reduction and anaesthetic gases (Perry et al. 2023). Searches were confined to the medical literature, and carbon emission data were not routinely reported for all the included primary studies. Keil et al. (2023) included LCAs which compared single-use and reusable healthcare products with similar functions (Keil et al. 2023). Interventions focused on non-invasive medical devices, inhalers, invasive medical devices and protective equipment. The review synthesis predominantly focused on GHG emission data, rather than carbon emissions, and did not consider the influence of individual specialities (Keil et al. 2023). Finally, the review conducted by Siu et al. (2017) compared the environmental impact of reusable versus disposable laparoscopic instruments (Siu et al. 2017). Searches for this review were limited to sources from the medical field, and the review authors did not conduct a quality appraisal of the included studies or report carbon-emission outcomes.

Two systematic reviews explored the environmental impact of telemedicine interventions in place of face-to-face patient care (Lange et al. 2022; Ravindrane and Patel 2022). The review by Ravindrane and Patel (2022) encompassed renal medicine, head and neck cancer, vascular surgery and urology specialities (Ravindrane and Patel 2022). While the review reported the impact of this type of intervention on carbon emissions, it did not consider variation in the use of telemedicine within different specialities (Ravindrane and Patel 2022). Lange et al. (2022) applied a transparency checklist for carbon footprint calculations within a systematic review of virtual care interventions (Lange et al. 2022). Overall, the review highlighted a saving of 148 kg carbon dioxide equivalents per patient, but indicated the evidence was weak, with the reported carbon footprint being highly heterogeneous (Lange et al. 2022). This review did not calculate contributions of individual specialities/pathways (Lange et al. 2022). In addition, these existing systematic reviews do not consider the evidence relating to the environmental impact of these interventions alongside the impact on patient and financial outcomes.

Why Is It Important to Carry Out Another Systematic Review and Develop an EGM?

Due to the lack of systematic reviews which consider carbon emissions associated with the patient pathway within individual specialities, further research is needed to enable the identification and transformation of the most carbon-intensive clinical pathways, while ensuring future models of care can be delivered in a cost-effective manner without increasing emissions or compromising patient care. A systematic review will enable the synthesis of existing evidence to address these research needs.

An EGM provides an overview of the quantity, quality and nature of primary evidence which already exists in this area (White et al. 2020), and how this evidence may help inform the patient care pathway. The interactive features of the EGM will enable users to identify and access evidence most suited to their needs.

Stakeholder Engagement

We have consulted with and worked closely alongside several stakeholder groups throughout the conduct of this review and the production of this review and EGM. Stakeholders included those requesting the review from the NHS England Greener NHS team, people with expertise in LCA methods or studies evaluating interventions to reduce carbon emissions within healthcare settings from the University of Exeter, and members of the PERSPEX patient and public involvement group. The method of engaging with and respective impact on the review process, of each of these stakeholder groups on the review process and outputs is summarised in Table 1.

Dimensions

Search Methods and Sources

Our review protocol was prospectively registered on the Open Repository Exeter (Shaw et al. 2023). The methods used to conduct and report the findings are consistent with the best practice approach for the conduct of systematic reviews and reporting of evidence synthesis (Centre for Reviews and Dissemination 2009; Page et al. 2021; Zumsteg et al. 2012). Below, we summarise how we identified relevant primary studies, quality appraised these and synthesised their findings.

The search strategy was developed by an information specialist (S.B.) in consultation with the review team and stakeholders (for further information, see ‘Stakeholder involvement’ section below). Our overall approach combined searches of bibliographic databases with backward and forward citation searches of studies that met the inclusion criteria, web searches of topically relevant organisations, searches of Google Search, and checking the included studies of topically similar systematic reviews. In addition, we inspected the Healthcare LCA database for relevant studies.

On 11th and 12th July 2023, we searched the following bibliographic databases with coverage of both health care and environmental science journals, including the health care databases MEDLINE and Embase (both via Ovid), the environmental science database Environment Complete (via EBSCO) and the multidisciplinary Science Citation Index database (via Web of Science, Clarivate Analytics). Searches of MEDLINE and Embase combined search terms for carbon emissions with search terms for relevant specialities (see MEDLINE search in Appendix 1, Search strategies). Medical speciality terms included generic terms for each speciality (e.g., gastroenterology, cardiology, etc.), diseases within each speciality which are treated in secondary care settings, and procedures within each speciality which are carried out in secondary care settings. A different approach was used to search Environment Complete and the Science Citation Index, which combined search terms for carbon emissions with generic terminology for hospital settings and secondary care (see Environment Complete search in Appendix 1, Search strategies). This approach was informed by our scoping searches, which suggested that potentially relevant studies published in environmental science journals typically use more generic terminology to describe medical settings than studies in medical and health care journals. A date limit of 2008 was applied across all databases. English language limits were applied where available.

The results of the bibliographic database searches were exported to EndNote 20 (Clarivate Analytics, Philadelphia, PA, USA) and de-duplicated using the automated de-duplication feature and manual checking.

We supplemented our bibliographic database searches by inspecting the HealthcareLCA database (https://healthcarelca.com/). This regularly updated resource indexes studies of LCAs of medical technologies and procedures, including carbon emissions, and can be filtered to identify studies relevant to specific medical specialities, including several specialities which are included in this review.

Forward and backward citation searches were conducted on all studies that met our inclusion criteria. Forward citation searching was carried out via the Science Citation Index (Web of Science, Clarivate Analytics) and Google Scholar (https://scholar.google.co.uk/), depending on which citation index indexed the relevant studies, which were identified. We also checked the included studies for any topically relevant systematic reviews that we identified during our scoping and screening processes.

We searched Google Search to identify studies not indexed in bibliographic databases or citation indexes, such as hospital-led evaluations published in grey literature format. Finally, we searched a selection of websites (see Appendix 1, Search strategies) for relevant studies.

The search strategies used for Google Search and websites are available in Appendix 1, Search strategies.

Data Collection and Analysis

Methods for Mapping

Description of Studies

Results of the Search

The bibliographic database searches identified 14,834 records. Following de-duplication, there were 11,826 unique records. At title and abstract screening, 11,571 records were excluded, leaving 255 to screen at full text. A further 951 records were identified via alternative search methods, including forwards citation chasing (n = 933), backwards citation chasing (n = 2), website searches (n = 13) and checking reference lists of relevant reviews (n = 3), of which 63 were sought for retrieval. Of the 305 full texts which could be retrieved, 213 were excluded for the reasons listed in Figure 1. Eighty-eight studies (92 articles) met eligibility criteria for inclusion in this review and are described in Supporting Material S1.

OPEN IN VIEWER

Figure 1

PRISMA diagram showing study selection process.

Overview of Included Studies

Of the 88 studies (92 articles) which met eligibility criteria for this review, the majority were published in peer-reviewed academic journals, aside from 13 published as non-peer-reviewed project web reports as part of carbon-reduction initiatives at individual NHS Trusts (Betts 2022; Bird 2022; Burton 2022; Chan 2023; Hardy 2022; Kodumuri et al. 2023; Kodumuri 2022; Langstaff 2023; Lewis et al. 2009; Materacki 2023; Milne 2010; Milne 2023; Nielsen 2022; Owens 2023). Thirty-three studies were conducted within the United Kingdom (Betts 2022; Bird 2022; Bond et al. 2009; Burton 2022; Chan 2023; Chuter et al. 2023; Connor, Lillywhite, et al. 2011; Connor, Mortimer, et al. 2011; Connor et al. 2019; Coombs et al. 2016; Cooper 2022; Cooper et al. 2023; Curtis et al. 2021; de Preux and Rizmie 2018; Dorrian et al. 2009; Hardy 2022; King et al. 2022; Kodumuri et al. 2023; Kodumuri 2022; Langstaff 2023; Lewis et al. 2009; Materacki 2023; Miah et al. 2019; Milne 2010; Milne 2023; Moussa et al. 2022; Moussa et al. 2021; Natale et al. 2022; Nielsen 2022; Owens 2023; Phull et al. 2023; Richards et al. 2022; Rizan and Bhutta 2022a; Udayaraj et al. 2019; Yong et al. 2022; Zander et al. 2011). Other countries included the United States (n = 19) (Beswick et al. 2016; Field et al. 2023; Frick et al. 2023; Jiang et al. 2021; Kemble et al. 2023; Le et al. 2022; Leapman et al. 2023; Meiklejohn et al. 2023; Mojdehbakhsh et al. 2021; Patel et al. 2023; Sherman et al. 2018; Sillcox, Blaustein, et al. 2023; Sillcox, Gitonga, et al. 2023; Thiel et al. 2015; Thiel et al. 2023; Thiel et al. 2018; Thota et al. 2020; Vo et al. 2023; Woods et al. 2015; Woolen et al. 2023), France (n = 5) (Baboudjian et al. 2022; Bendine et al. 2020; Chambrin et al. 2023; Filfilan et al. 2021; Rouviere et al. 2022), Germany (n = 6) (Arndt et al. 2023; Buttner et al. 2021; Klein 2023; Leiden et al. 2020; Muschol et al. 2022; Schulte et al. 2021), Australia (n = 4) (Andrew et al. 2020; Davis et al. 2018; McAlister et al. 2022; Wombwell et al. 2023), Canada (n = 4) (Cheung et al. 2023; Forner et al. 2021; Lambert et al. 2023; Tselapedi-Sekeitto 2023), Sweden (n = 3) (Boberg et al. 2022; Holmner et al. 2014; Stripple et al. 2008), Ireland (n = 3) (Croghan et al. 2021; Hogan et al. 2022; McCarthy et al. 2014), multiple countries (n = 2) (Asghari and Al-e-Hashem 2020; Vaidya et al. 2022), and one study was conducted within Austria (Winklmair et al. 2023), China (Chen et al. 2017), Denmark (Sørensen and Grüttner 2018), Italy (Fuschi et al. 2023), New Zealand (McLachlan et al. 2021), Portugal (Neves et al. 2022), Spain (Lopez-Munoz et al. 2023), Switzerland (Heye et al. 2023).

Twenty-eight studies used LCA-informed methods to calculate carbon emissions. Six studies conducted inventory analysis (Holmner et al. 2014; Kemble et al. 2023; Sillcox, Gitonga, et al. 2023; Thiel et al. 2018; Winklmair et al. 2023; Wombwell et al. 2023), or used methods informed by a LCA approach (n = 3) (Chuter et al. 2023; Connor, Lillywhite, et al. 2011; de Ridder et al. 2022), such as healthcare sustainability mode and effect analysis (de Ridder et al. 2022), or component analysis (Connor, Lillywhite, et al. 2011). Nineteen studies used a full LCA approach (Baboudjian et al. 2022; Boberg et al. 2022; Davis et al. 2018; Fuschi et al. 2023; Hogan et al. 2022; Le et al. 2022; Leapman et al. 2023; Leiden et al. 2020; Lopez-Munoz et al. 2023; McAlister et al. 2022; Meiklejohn et al. 2023; Rizan and Bhutta 2022a; Rouviere et al. 2022; Schulte et al. 2021; Sherman et al. 2018; Sørensen and Grüttner 2018; Stripple et al. 2008; Thiel et al. 2015; Thiel et al. 2023). Of the 35 studies conducted within the United Kingdom, one used full LCA methodology (Rizan and Bhutta 2022a).

Of the studies which did not use LCA methods to calculate carbon emissions, 19 were based on an experimental study design, including three RCTs (Coombs et al. 2016; Field et al. 2023; Muschol et al. 2022), five controlled trials (Chen et al. 2017; Cooper et al. 2023; Frick et al. 2023; Hogan et al. 2022; Tselapedi-Sekeitto 2023), one feasibility study (Beswick et al. 2016), and 11 before and after studies (Arndt et al. 2023; Betts 2022; Cheung et al. 2023; Kodumuri et al. 2023; Kodumuri 2022; Langstaff 2023; Materacki 2023; McCarthy et al. 2014; McLachlan et al. 2021; Neves et al. 2022; Nielsen 2022; Udayaraj et al. 2019). Nine were modelling studies (Asghari and Al-e-Hashem 2020; Bird 2022; Burton 2022; de Preux and Rizmie 2018; Hardy 2022; Milne 2010; Milne 2023; Owens 2023; Zander et al. 2011). The remaining studies used observational methods, with the most common study designs being cross-sectional (n = 9) (Buttner et al. 2021; Connor, Mortimer, et al. 2011; Croghan et al. 2021; Filfilan et al. 2021; Forner et al. 2021; King et al. 2022; Lewis et al. 2009; Mojdehbakhsh et al. 2021; Patel et al. 2023), retrospective or prospective cohort (n = 16) (Bendine et al. 2020; Beswick et al. 2016; Chambrin et al. 2023; Connor et al. 2019; Curtis et al. 2021; Dorrian et al. 2009; Heye et al. 2023; Jiang et al. 2021; Miah et al. 2019; Natale et al. 2022; Richards et al. 2022; Thota et al. 2020; Vaidya et al. 2022; Vo et al. 2023; Woolen et al. 2023; Yong et al. 2022), and database review (n = 6) (Andrew et al. 2020; Bond et al. 2009; Lambert et al. 2023; Phull et al. 2023; Sillcox, Blaustein, et al. 2023; Woods et al. 2015). Further detail regarding the methods used in the studies based on LCA is provided as Supporting Material S2.

Studies were classified according to the five broad intervention categories: ‘Accessing care’ (n = 29) (Andrew et al. 2020; Arndt et al. 2023; Asghari and Al-e-Hashem 2020; Beswick et al. 2016; Bond et al. 2009; Connor, Mortimer, et al. 2011; Connor et al. 2019; Croghan et al. 2021; Curtis et al. 2021; Dorrian et al. 2009; Filfilan et al. 2021; Forner et al. 2021; Holmner et al. 2014; Jiang et al. 2021; King et al. 2022; Lambert et al. 2023; Lewis et al. 2009; McLachlan et al. 2021; Miah et al. 2019; Mojdehbakhsh et al. 2021; Muschol et al. 2022; Natale et al. 2022; Patel et al. 2023; Richards et al. 2022; Sillcox, Blaustein, et al. 2023; Sillcox, Gitonga, et al. 2023; Thiel et al. 2023; Thota et al. 2020; Tselapedi-Sekeitto 2023; Udayaraj et al. 2019), ‘Setting’ (n = 20) (Betts 2022; Bird 2022; Buttner et al. 2021; Chambrin et al. 2023; de Ridder et al. 2022; Heye et al. 2023; Klein 2023; Kodumuri et al. 2023; Kodumuri 2022; Lopez-Munoz et al. 2023; McAlister et al. 2022; McCarthy et al. 2014; Milne 2010; Milne 2023; Neves et al. 2022; Owens 2023; Vo et al. 2023; Winklmair et al. 2023; Woolen et al. 2023; Yong et al. 2022), ‘Product level’ (n = 16) (Baboudjian et al. 2022; Boberg et al. 2022; Chan 2023; Davis et al. 2018; Field et al. 2023; Hogan et al. 2022; Kemble et al. 2023; Le et al. 2022; Leiden et al. 2020; Moussa et al. 2022; Moussa et al. 2021; Rizan and Bhutta 2022a; Schulte et al. 2021; Sherman et al. 2018; Sørensen and Grüttner 2018; Stripple et al. 2008; Wombwell et al. 2023), ‘Care delivery’ (n = 16) (Burton 2022; Chen et al. 2017; Connor, Lillywhite, et al. 2011; Coombs et al. 2016; Cooper 2022; Cooper et al. 2023; de Preux and Rizmie 2018; Frick et al. 2023; Fuschi et al. 2023; Langstaff 2023; Leapman et al. 2023; Meiklejohn et al. 2023; Nielsen 2022; Phull et al. 2023; Thiel et al. 2015; Vaidya et al. 2022; Woods et al. 2015; Zander et al. 2011), and ‘Multiple components’ (n = 7) (Bendine et al. 2020; Cheung et al. 2023; Chuter et al. 2023; Hardy 2022; Materacki 2023; Rouviere et al. 2022; Thiel et al. 2018). Urology (n = 14) (Baboudjian et al. 2022; Connor et al. 2019; Croghan et al. 2021; Davis et al. 2018; Filfilan et al. 2021; Fuschi et al. 2023; Hogan et al. 2022; Kemble et al. 2023; Leapman et al. 2023; Miah et al. 2019; Natale et al. 2022; Phull et al. 2023; Stripple et al. 2008; Wombwell et al. 2023), Gastroenterology (n = 12) (Betts 2022; Boberg et al. 2022; King et al. 2022; Le et al. 2022; Lopez-Munoz et al. 2023; Materacki 2023; Neves et al. 2022; Owens 2023; Rizan and Bhutta 2022a; Sherman et al. 2018; Sillcox, Blaustein, et al. 2023; Sillcox, Gitonga, et al. 2023; Yong et al. 2022), and Oncology/Radiation oncology (n = 13) were the most common specialities represented (Beswick et al. 2016; Cheung et al. 2023; Coombs et al. 2016; Forner et al. 2021; Frick et al. 2023; Jiang et al. 2021; Lambert et al. 2023; Langstaff 2023; Lewis et al. 2009; Patel et al. 2023; Thota et al. 2020; Vaidya et al. 2022; Woods et al. 2015). The most common intervention evaluated was telemedicine (n = 26) (Andrew et al. 2020; Arndt et al. 2023; Beswick et al. 2016; Connor, Mortimer, et al. 2011; Connor et al. 2019; Croghan et al. 2021; Curtis et al. 2021; Dorrian et al. 2009; Filfilan et al. 2021; Holmner et al. 2014; Jiang et al. 2021; King et al. 2022; Lambert et al. 2023; Lewis et al. 2009; McLachlan et al. 2021; Miah et al. 2019; Mojdehbakhsh et al. 2021; Muschol et al. 2022; Natale et al. 2022; Patel et al. 2023; Richards et al. 2022; Sillcox, Blaustein, et al. 2023; Sillcox, Gitonga, et al. 2023; Thiel et al. 2023; Thota et al. 2020; Tselapedi-Sekeitto 2023; Udayaraj et al. 2019), which included three studies using LCA methods (Holmner et al. 2014; Sillcox, Gitonga, et al. 2023; Thiel et al. 2023). LCA methods were most commonly used to evaluate interventions within the ‘Product level’ category, specifically interventions comparing carbon emissions associated with reuseable versus single use equipment (n = 12) (Baboudjian et al. 2022; Boberg et al. 2022; Davis et al. 2018; Hogan et al. 2022; Kemble et al. 2023; Le et al. 2022; Leiden et al. 2020; Rizan and Bhutta 2022a; Schulte et al. 2021; Sherman et al. 2018; Sørensen and Grüttner 2018; Wombwell et al. 2023), with the highest number in urology product-level interventions (n = 6) (Baboudjian et al. 2022; Davis et al. 2018; Hogan et al. 2022; Kemble et al. 2023; Stripple et al. 2008; Wombwell et al. 2023). Further detail regarding interventions evaluated by the studies using LCA methods is provided in Appendix 5, Description of interventions evaluated by LCA studies.

Quality Appraisal of Included Studies

The quality appraisal scores of LCA studies ranged from 11.5 to 34 (see Table 2). The majority of the studies had a clear aim or rationale and were clear on the paper's intended application and/or audience. All but three studies stated the lifecycle assessment method clearly (Chuter et al. 2023; Hogan et al. 2022; Rouviere et al. 2022). However, only eight studies explicitly reported that they had conducted the study in accordance with ISO standards (ISO 14040 series) (Boberg et al. 2022; Fuschi et al. 2023; Leapman et al. 2023; McAlister et al. 2022; Rizan and Bhutta 2022a; Schulte et al. 2021; Sherman et al. 2018; Thiel et al. 2023).

Not all studies were classified as full LCAs with some stating that the study was a ‘simplified’ LCA (Hogan et al. 2022; Holmner et al. 2014), others were inventory analyses as they did not consider environmental impacts beyond carbon emissions (Chuter et al. 2023; Davis et al. 2018; Fuschi et al. 2023; Hogan et al. 2022; Kemble et al. 2023; Leapman et al. 2023; McAlister et al. 2022; Sillcox, Gitonga, et al. 2023; Winklmair et al. 2023; Wombwell et al. 2023). Only five studies reported full details on the functional unit (Baboudjian et al. 2022; Boberg et al. 2022; Holmner et al. 2014; Schulte et al. 2021; Stripple et al. 2008), while nine did not report any details at all (Davis et al. 2018; de Ridder et al. 2022; Fuschi et al. 2023; Hogan et al. 2022; Kemble et al. 2023; Lopez-Munoz et al. 2023; Thiel et al. 2018; Winklmair et al. 2023; Wombwell et al. 2023). Thirteen studies reported details on the systems studied and defined the system boundaries, often described as ‘cradle to grave’ (Baboudjian et al. 2022; Connor, Mortimer, et al. 2011; Fuschi et al. 2023; Holmner et al. 2014; Kemble et al. 2023; Leapman et al. 2023; Lopez-Munoz et al. 2023; McAlister et al. 2022; Rizan and Bhutta 2022a; Schulte et al. 2021; Sillcox, Gitonga, et al. 2023; Stripple et al. 2008; Thiel et al. 2023). Just over a third of the studies fully reported details on the system covered – production, use/reuse and disposal of materials and energy (Connor, Lillywhite, et al. 2011; Fuschi et al. 2023; Holmner et al. 2014; Leapman et al. 2023; McAlister et al. 2022; Meiklejohn et al. 2023; Schulte et al. 2021; Sillcox, Gitonga, et al. 2023; Stripple et al. 2008; Thiel et al. 2023). More than half the studies (n = 17) fully explained the data collection process, the data included, and the source(s) of reference data which was most commonly the Ecoinvent database (Boberg et al. 2022; Connor, Lillywhite, et al. 2011; Fuschi et al. 2023; Holmner et al. 2014; Kemble et al. 2023; Le et al. 2022; Leapman et al. 2023; Lopez-Munoz et al. 2023; McAlister et al. 2022; Meiklejohn et al. 2023; Rizan and Bhutta 2022a; Schulte et al. 2021; Sherman et al. 2018; Sillcox, Gitonga, et al. 2023; Sørensen and Grüttner 2018; Thiel et al. 2015; Thiel et al. 2023). Authors could report obtaining detailed data from manufacturers (e.g., Hogan et al. 2022), and others noted a lack of data on the material composition of devices (e.g., Baboudjian et al. 2022; Le et al. 2022). Allocation was not always necessary (Chuter et al. 2023; Davis et al. 2018; Holmner et al. 2014; Sillcox, Gitonga, et al. 2023; Thiel et al. 2023; Winklmair et al. 2023), but where environmental exchanges had to be allocated to different products, a number of studies described and justified the allocation (Boberg et al. 2022; Connor, Lillywhite, et al. 2011; de Ridder et al. 2022; Lopez-Munoz et al. 2023; Rizan and Bhutta 2022a; Schulte et al. 2021; Sherman et al. 2018; Thiel et al. 2015; Thiel et al. 2018; Wombwell et al. 2023). Only five studies scored highly on reporting impact assessment (Baboudjian et al. 2022; Fuschi et al. 2023; Rizan and Bhutta 2022a; Schulte et al. 2021; Thiel et al. 2015). Some reported the tools used for assessing environmental impacts: (i) the US EPA's impact assessment model, TRACI (Tool for the Reduction and Assessment of Chemical and other Environmental Impacts) (Leapman et al. 2023; Meiklejohn et al. 2023; Sherman et al. 2018; Thiel et al. 2015; Thiel et al. 2023; Thiel et al. 2018); (ii) ReCiPe Midpoint Hierarchist model (Le et al. 2022; Leiden et al. 2020; McAlister et al. 2022; Rizan and Bhutta 2022a); and (iii) the Eco-Indicator 99 tool (Stripple et al. 2008).

Only three studies used both uncertainty analyses and sensitivity analyses to contextualise their results (Boberg et al. 2022; Holmner et al. 2014; Thiel et al. 2023). For the majority of studies (n = 23), their conclusions were consistent with their goals and supported by their results. In total, three of the 19 LCA studies (15.8%) report 95% CI data (Baboudjian et al. 2022; Lopez-Munoz et al. 2023; Thiel et al. 2015). Three studies did not disclose a potential conflict of interest or include a funding declaration (Connor, Lillywhite, et al. 2011; Meiklejohn et al. 2023; Thiel et al. 2018). Five studies made a funding declaration but asserted that receiving funding from manufacturing companies such as Amu A/S and Neo Medical S. A. had not influenced the results of studies (Baboudjian et al. 2022; Leiden et al. 2020; Rizan and Bhutta 2022a; Sørensen and Grüttner 2018; Stripple et al. 2008). One study reported a research agreement with an industry partner (Vanguard AG, Germany) who provided primary data for the ‘medical remanufacturing route’ and the ‘bill of materials’ for the catheter under investigation (Schulte et al. 2021).

Summary of Main Findings: EGM

This EGM provides a brief description of where primary research exists across the patient pathway for 13 specialities in secondary healthcare. Further details regarding intervention findings specific to individual specialities can be found in Appendix 6, Speciality-specific findings and link to online interactive EGM.

Narrative Synthesis

This section presents the narrative synthesis of all included studies, grouped into five broad intervention categories: Accessing care (n = 29), Setting (n = 19), Product level (n = 16), Care delivery (n = 16) and Multiple components (n = 7). Within each of these categories, studies are grouped into those using LCA-informed methods and those that use non-LCA methods. Within each of these methodological groups, studies evaluating similar interventions were clustered together to explore key outcomes such as carbon emissions, patient outcomes and service costs.

A summary providing an overview of findings from each intervention category is provided at the start of each section. Please see Supporting Materials S3 and S4 for details regarding methods and specific outcomes from individual studies.

Summary of Main Results

This systematic review aimed to examine the effectiveness of interventions in reducing the carbon footprint within medical specialities in secondary healthcare and summarise how this evidence could inform the patient care pathway. Eighty-eight studies met the eligibility criteria. We presented the evidence in an EGM, structured according to a secondary healthcare patient care pathway. Urology (n = 14), gastroenterology (n = 12). oncology/radiation oncology (n = 13) and renal (n = 11) were the most common specialities represented. Across different specialities, the majority of evidence was found in the first three stages of the patient care pathway (Initial assessment/diagnostic tests, initial treatment or follow-up). The exception to this was the renal speciality, where most of the evidence was within the ‘ongoing care’ segment of the patient care pathway. This may reflect the ongoing treatment required by individuals receiving haemodialysis and thus the associated opportunities to conserve energy and resources. There was limited evidence within the ‘discharge’ segment of the care pathway across all specialities. Evidence relating to the wider healthcare setting was clustered within the gastroenterology (n = 5) and radiology specialities (n = 5).

Twenty-eight studies used LCA-informed methodology. Nine LCA studies were appraised as Low risk of bias (Baboudjian et al. 2022; Boberg et al. 2022; Holmner et al. 2014; McAlister et al. 2022; Rizan and Bhutta 2022; Schulte et al. 2021; Sherman et al. 2018; Thiel et al. 2015, 2023), 14 as Medium risk of bias (Chuter et al. 2023; Connor, Lillywhite, et al. 2011; de Ridder et al. 2022; Fuschi et al. 2023; Kemble et al. 2023; Le et al. 2022; Leapman et al. 2023; Leiden et al. 2020; Lopez-Munoz et al. 2023; Meiklejohn et al. 2023; Sillcox, Gitonga, et al. 2023; Sørensen and Grüttner 2018; Stripple et al. 2008; Wombwell et al. 2023), and 5 as High risk of bias using the criteria detailed within the methods section above (Davis et al. 2018; Hogan et al. 2022; Rouviere et al. 2022; Thiel et al. 2018; Winklmair et al. 2023). Three of the five studies appraised as High risk of bias were inventory analysis (Davis et al. 2018; Hogan et al. 2022; Winklmair et al. 2023). Study characteristics associated with a High risk of bias appraisal rating included the following issues regarding the goal and scope of the studies: poor definition of functional unit used, poor description and/or justification of system boundaries and life cycle stages included in the analysis, calculations did not consider production, use/reuse and disposal of materials and energy. In terms of the inventory analysis, High risk of bias scores were associated with poor description of impact categories, categorisation method and software used and poor reporting of results in the context of the functional unit. Regarding interpretation of findings, studies appraised as High risk of bias did not contextualise their findings through the use of uncertainty and/or sensitivity analysis.

Interventions evaluated by the included studies were classified into one of five broad categories: ‘Accessing care’ (n = 29), ‘Setting’ (n = 20), ‘Product level’ (n = 16), ‘Care delivery’ (n = 16) and ‘Multiple components’ (n = 7). The two largest groups of evidence were for studies evaluating telehealth (n = 26) and reuseable equipment (n = 13) interventions. Telehealth interventions were predominantly evaluated using non-LCA methodology (n = 23) and while carbon emissions favoured telemedicine interventions when compared to face-to-face care, these calculations often only considered patient-travel saved and did not account for carbon emissions associated with the use of the digital pathway or other parts of the patient care pathway, such as the impact on primary care. In general, the majority of patient and cost outcomes evaluated favoured the telemedicine intervention, although most outcomes were based on descriptive or narrative analyses. These findings were reflected in the systematic review by Ravindrane and Patel (2022), which explored the environmental impact of telemedicine instead of face-to-face care in healthcare. They highlighted that the benefit of telemedicine in terms of carbon emission reduction was dependent on energy consumption of the telemedicine systems, number of patients, mode of transport used, and distance of travel avoided and indicated that improvements to modelling used within studies were needed, including the use of sensitivity analysis and transparent reporting of assumptions used (Ravindrane and Patel 2022). Lange et al. (2022) also highlighted the poor methodological quality of carbon emission methodology used in their review of telemedicine interventions within healthcare (Lange et al. 2022).

Interventions comparing carbon emissions associated with the use of reuseable versus disposable surgical equipment represented the largest group of studies utilising LCA methods. For studies within the gastroenterology speciality, reuseable equipment was associated with reduced carbon emissions. Within urology, this finding was reversed, with disposable instruments found to be associated with reduced carbon emissions. However, despite the quality of these studies being appraised as mainly High or Medium within this review, questions regarding the accuracy of use of characterisation factors, quantity of materials used in disposable versus reuseable equipment packs and how carbon emissions were assigned to the reprocessing stage of reuseable equipment mean confidence in this finding is uncertain (Brighton and Sussex Medical School, Centre for Sustainable Healthcare, and UK Health Alliance on Climate Change 2023; Rizan and Bhutta 2022). The latter finding contrasts with findings from two other systematic reviews, which indicate reuseable devices are associated with improved environmental outcomes (Drew et al. 2021; Siu et al. 2017), although limitations to the evidence base include methodological heterogeneity and lack of background life cycle inventory data for surgical inputs (Drew et al. 2021), and lack of cost-comparison studies (Siu et al. 2017). The uncertainty regarding the beneficial effects of reuseable equipment on carbon emissions within urology arising from this review underscores the importance of considering the full product pathway within an LCA approach and ensuring the system boundaries for the change being considered reflect all parts of the patient care pathway and product life-cycle. The composition of products evaluated and processes associated with, for example, transport, reprocessing of reuseable devices and waste disposal are highly context dependent, with alterations to these processes potentially having a huge impact on estimated carbon emission calculations (Hogan and Hennessey 2023; Leiden et al. 2020). Thus, it can be challenging to generalise findings across LCA studies, even when conducted in similar countries/health systems for the same type of intervention, and emphasises the importance of incorporating sensitivity analysis into LCAs. It also highlights the importance of considering how to reduce carbon emissions associated with the processes supporting the manufacture, transport and reprocessing of disposable and/or reuseable equipment as a target for future interventions. This is an alternative focus to comparing emissions associated with disposable versus reuseable equipment and promotes addressing carbon emissions associated with known ‘hot spots’ in the lifecycle of both types of product, such as manufacturing for disposable products and reprocessing for reuseable products. This may complement recommendations from the Green Surgery and MedTech Circular Economy reports to, among other actions, pursue the use of reuseable equipment to reduce carbon emissions and overcome challenges within supply chains, resource scarcity, healthcare disparities and waste production (Brighton and Sussex Medical School, Centre for Sustainable Healthcare, and UK Health Alliance on Climate Change 2023).

Finally, while waste management/reduction interventions were associated with reduced carbon emissions (n = 12), interventions were highly heterogeneous with limited consideration of patient or cost outcomes. Eight non-LCA studies found reduced carbon emissions were associated with energy conservation interventions, the majority of which were conducted within radiology/radiotherapy settings and focused on the impact (or potential impact) of turning machines off when not in use.

Strengths and Limitations of This Review

We have conducted a comprehensive systematic review of the literature, which identifies and synthesises comparative studies evaluating interventions to reduce carbon emissions across nine specialities within secondary healthcare. We grouped these studies by broad intervention category to enable identification of carbon emission, patient, and cost outcomes relevant to specific interventions within each speciality, separating out evidence from studies which used LCA methods to calculate carbon emissions to highlight findings supported by the most methodologically robust evidence base. Our synthesis was informed by the critical appraisal of these LCA studies. Due to a lack of standardised appraisal tools and associated risk of bias thresholds, the critical appraisal scores awarded by reviewers were calibrated based on the set of studies included in the review. Further work to establish the validity of these thresholds is required. Unfortunately, there was a high degree of heterogeneity between types of intervention conducted within individual specialities, which made it challenging to identify interventions which were effective in reducing carbon emissions within similar and across different contexts. While any intervention with a direct link to the specialities indicated was eligible for inclusion in this review, there may be other interventions, such as the frequency with which blood pressure cuffs are sanitised (Sanchez et al. 2020), which may be relevant across multiple specialities but have been excluded from this review, as this has not been clearly stated. Thus, the evidence in this review may not reflect the full range of interventions available to clinicians within a particular speciality.

The number of studies including patient and cost outcomes alongside carbon emission calculations was also limited. This may reflect our inclusion criteria, which required studies to measure carbon emissions. Thus, unless related to one of our included studies, studies purely focused on patient outcomes or service costs would have been excluded. The paucity of studies reporting patient clinical outcomes and satisfaction from both intervention and control groups may also reflect that studies using a before-and-after design or conducting a retrospective database review relied on data recorded on electronic databases, where these outcomes may not be routinely recorded. Our inclusion criteria also required that included studies referenced a particular speciality, which may have resulted in the exclusion of otherwise relevant interventions, particularly within the ‘systemic interventions’ section of the care pathway.

Within studies drawing on an LCA approach, the lack of transparency in the reporting of methodological details raised issues of comparability and generalisability. The variability among LCA studies may be explained by data collection and calculation procedures, along with the researchers' assumptions and choices of background inventory databases. Although the Ecoinvent database was most commonly reported, there was a wide variety of secondary sources used by researchers, many of which were originally compiled for other purposes, ranging from government documents to other research papers and conference proceedings. In some instances, researchers reported using manufacturer details to calculate raw material composition of devices (e.g., Hogan et al. 2022; Kemble et al. 2023; Rizan and Bhutta 2022a, 2022b), while others could not access such data and based their calculations on available data for similar devices (e.g., Le et al. 2022). In addition, comparability was hindered by the lack of consistency in how studies defined and reported the system boundaries for the individual LCA studies. The difficulty in generalising results from the LCA was that the data could be specific to a particular context or intervention. The geographical setting of a study was important, particularly in relation to calculating the electricity supply. Studies undertaken in the United States, for example, assumed a US electricity supply with sources derived largely from fossil fuel (e.g., Leapman et al. 2023), and results based on these assumptions were not likely to apply to other countries with cleaner energy sources such as Sweden (e.g., Holmner et al. 2014).

Carbon emission calculations used within non-LCA studies were typically narrow in scope, focusing on the use and/or reuse of products, with less consideration of other factors within the wider system, which may also influence carbon emissions of the intervention, for example, energy used by both health services and patients. The extent to which carbon-emission calculations in non-LCA studies considered emissions associated with the manufacture of equipment, vehicles or fuel, transport and/or waste management was also limited and dependent on the intervention in question. These issues were particularly evident for interventions such as telemedicine or remote delivery of care, where carbon-emission calculations were typically based on non-statistical comparisons of patient travel distance saved because of reduced number of visits to hospital, with less consideration of factors such as staff and patient energy use via heating, lighting and/or internet access. For studies focusing on waste management/recycling initiatives as part of local initiatives to reduce carbon emissions within specific NHS trusts, carbon-emission calculations rarely consider emissions associated with the transport and recycling of waste, which would otherwise have been destroyed. However, where comparable interventions existed between the two groups of evidence, findings from non-LCA studies generally reflected those in LCA studies.

We have presented all included studies within an interactive EGM, which displays the evidence relative to the patient care pathway for each speciality. This will enable evidence users to locate evidence relevant to their interests and requirements, and highlight where groups of evidence and gaps exist. The smaller quantity of evidence relating to the ‘Discharge’ part of the patient care pathway is likely influenced by the inclusion criteria for this review, which focused on interventions led by secondary healthcare. Thus, interventions such as self-management or ongoing support within the community would not have been captured in the EGM. In addition, most interventions included in this review, and thus within the patient care pathway in the EGM, often reflect tangible changes that clinicians can make during their everyday practices at key patient care pathway. While more complex interventions, which changed the structure of patient care pathways and multi-component interventions, were represented within this map, they were in the minority. This may reflect the increased difficulty in implementing and evaluating these types of intervention. Thus, while some of the interventions included in this review may correspond with known carbon hotspots (e.g., those targeting use of resources within surgical settings), others may indicate situations where clinicians feel they can quickly implement a change to address an area of high resource/energy consumption within their local service context.

Implications for Research

–

Existing research relating to carbon emissions reflects a narrow range of all the possible interventions/specialities available. Further research is needed to fill the gaps highlighted in the EGM, particularly evidence relating to the ongoing care or discharge of patients or relating to obstetric and respiratory specialities.

Implications for Policy

–

Our EGM provides a resource to identify where gaps in primary evidence exist on the patient care pathway, both within and across different specialities, making it a useful tool to inform commissioning of future research.

Implications for Practice

–

There is tentative evidence to indicate that interventions which reduce the distance patients need to travel to access care are associated with reduced carbon emissions. However, the impact on patient clinical outcomes and patient satisfaction is inconclusive and further research which addresses the methodological limitations highlighted above is required to increase confidence in reported findings.