The primary purpose of this paper was to systematically review how the carbon footprint of telemedicine can be accounted for. Additionally, we propose a preliminary but practical calculator to roughly estimate emissions of an individual health care supplier based on the results and our example calculation in eyecare. We accounted for the reported remote treatment of 51,028 patients, resulting in 13,318,882 km saved. The median visit distance was 131 km (IQR:52–386) with a median emission of 26.3 kgCO2/visit (IQR:10.6–94.4). The average emission by ground travel without LCA is 0.256 kgCO2/km (SE = 0.003), while it is 0.250 kgCO2/km (SE = 0.009) including LCA. Telemedicine becomes carbon cost-effective when the roundtrip distance is greater than 12.2 km. Time savings for patient traveling had a wide range depending on the country with a median of 2.8 hours (IQR:1.7–4.1). The CCrC model is a preliminary and simplified attempt to roughly capture the individual carbon footprint savings of telemedicine, furthering the debate on how to structurally measure and calculate CO2 emissions in telemedicine.
This review highlights that travel distance by itself is a rough marker on estimating the total emissions as it is the biggest individual contributor. However, results show that distances vary widely across countries and transportation types. Travel emissions are region-specific, and the average emissions of a telemedicine intervention in one place can differ significantly from the same intervention in another. Literature on additional variables is limited and often unsystematically evaluated.
Although travel distance is the biggest contributor, only including travel emissions for carbon footprint comparison is incomplete and arguably underestimates the true carbon impact. However, a very limited number of studies have been conducted that account for streamlined LCAs, and approaching an adequate estimation raises some concerns. Our findings show that by not including external factors such as a streamlined LCA, the actual saved emissions are structurally overestimated. This result may be counterintuitive, but it is a direct result of the authors’ freedom to consider only the variables they consider necessary, which often resulted in weighing the carbon footprint of telemedicine without accounting for emissions related to in-person clinic. As an example, Masino et al. solely included emissions of a computer screen contrary to Bartlett et al. who considered LCA variables that are not related to telemedicine such as staff travel and overhead. Authors should consider using a systematic approach unless they have valid reasons not to do so for better precision of a streamlined LCA like suggested by Lange et al.(55) Calculations also vary widely due to differences in efficiency of product manufacturing, internet speed, energy use, and transport emission rates.(17, 20, 45, 46, 56, 57) Although solvable, these uncertainties increase the difficulty of estimating and comparing LCAs. Since we have underlined the relevance of estimating LCAs, we recommend further studies in this direction to increase overall precision and generalizability.
Staff travel is investigated incidentally. It represents a vast quantity of emissions as health care is a labor-intensive industry and rapidly increases their contribution to emissions, making it the primary contributor to the footprint of virtual consultations and the second largest contributor to the footprint of face-to-face consultations. Although these emissions are crucial to analyze, staff travel is difficult to decrease in an existing model of health care delivery, where professionals often provide physical examination to multiple patients daily (i.e., mixed with teleconsultations).
To interpret the sum of all included variations on the scale of an individual clinic, we developed a simplified tool that allows individual users to customize inputs and estimate the expected CO2 savings of telemedicine versus usual care based on the results of this paper (Supplementary File 2). For example, replacing 67% of post-operative cataract surgery visits with remote visits would result in savings of 471,000 kgCO2/year from reduced car travel alone. Correcting for LCA results in an estimated savings of 459,000 kgCO2/year, which is equivalent to the average yearly electricity consumption of almost 300,000 Dutch households, or more than 76,500 cataract surgeries in India.(58, 59) While an overestimation by 2.5% when LCA is excluded may seem negligible, the difference of 12,000 kgCO2 given in this example represents a substantial discrepancy in absolute terms for large-scale calculations.
Applying the CCrC to the papers included in our ground transportation analysis demonstrated that the median ratio of the calculated value to the originally reported values is 0.92 (IQR:0.8–0.98). Values especially deviate from 1 when the source used in the original reporting differs greatly from what may be considered more appropriate now. For example, Andrew et al. used US EPA values for their study based in Australia, and Holmner et al. referenced papers about global estimates for vehicle fuel efficiency when their study was based in one of the countries with the lowest estimate for emissions per unit distance. Discrepancies can be further explained by gradually increased vehicle fuel efficiency, which would result in a lower estimate for emissions closer to the present day compared to several years ago. If we only consider articles published from 2015 onwards, the median ratio of the calculated value to the originally reported values increases to 0.96 (IQR:0.89–1.01). Residual discrepancies may be attributed to vehicle fuel efficiency improvement even in recent years as well as differences in LCA accounting.
There are limitations to this paper, such as the ambiguous interpretation of ‘streamlined’ LCA’s, which lead to substantial variability in LCAs between articles. Additionally, the number of studies available limits the LCA-inclusive estimate for telemedicine emissions. Attempts to estimate emissions worldwide using various sources of CO2 emission per region may also be inconsistent depending on how frequently such reports are updated.
The strength of the CCrC calculation lies in its conceptualization as a practical calculator that provides insight into emissions of the individual health care supplier. With that strength comes its limitation; we acknowledge that the CCrC model is simplified and preliminary, for example in its generalization of emissions for petrol and diesel vehicles. However, the included references represent a weighted average of these emissions on a per-country basis, and the “Custom Case” functionality allows users to fine-tune the estimate based on the vehicle type of interest amongst other parameters. The references would also need to be updated regularly to accurately reflect the current state of such factors as vehicle fuel efficiency and vehicle usage. This furthers the debate on how the structural measurement and calculation of CO2 emission in telemedicine can be improved for practical usage by practitioners.
There is a change in thinking on the role of carbon footprint from a societal perspective. Societal gains are usually researched in a Health Technology Assessment (HTA). However, Carbon footprint typically falls outside the scope of HTA research, despite the impact of CO2 emissions on society.(60, 61) GHGs cause significant harm to population health, which means that carbon footprint should be taken into account to reflect everyone’s responsibility for climate change.(62–64) Nonetheless, it is a challenge to define the place of carbon footprint in HTA research. Usually, costs are compared to effects or utilities, e.g., quality-adjusted life years. Difficulties arise as kg’s of emission cannot be directly included in conventional studies. A suggested solution to weigh carbon emission would be to use Multi Criteria Decision Analyses (MCDA), a form of HTA. It is known for its ability to capture multiple and conflicting criteria, measured in different units within a single overall estimate of value.(65, 66) We encourage researchers that aim to include the carbon footprint in the societal value of an intervention to use the MCDA.
Apart from the health-environmental impact, other reasons for implementation of telemedicine are a prohibitive length or complexity of travel (e.g., in rural areas or disaster management), when public transport is too difficult (e.g., large cities or elderly populations), or for social distancing during the COVID-19 pandemic. Additional considerations include cost savings and broadening access to care.(67),(68) Moreover, digitalization is often mentioned as a possible solution for healthcare challenges in the coming decades.(69, 70) From a future perspective, one could imagine new hospitals being smaller due to increased digital care, potentially becoming completely digital with examples such as Kysos and Mobile Doctors.(71, 72)